HTTP

First, go to HTTP in the top navigation bar and then click Export.

Click the button to download a ZIP file that contains all of the HTTP mappings in the WireMock format.

You can also export programmatically using the Traffic Parrot Management APIs.

As well as the native WireMock-format ZIP, Traffic Parrot can export your HTTP mappings as an OpenAPI 3.0 document. This is the inverse of importing an OpenAPI specification: the mappings you have recorded or built are translated back out into a standard OpenAPI document you can share, diff, or feed into other tooling.

On the Export page (HTTP in the top navigation bar, then Export), alongside the Download Mappings button that downloads the native WireMock-format ZIP, there is an Export as OpenAPI (YAML) button. Click it to download the currently active scenario's mappings as an OpenAPI 3.0 YAML document, with an inferred response schema alongside each example and the coverage report described below embedded in the document. Unlike the native ZIP, which is a round-trippable archive you can re-import unchanged, the OpenAPI document is a standard contract for sharing and tooling.

The OpenAPI export is served off the same export endpoint as the native ZIP by adding the ?format=openapi query parameter, so you can also trigger it programmatically. Without the parameter the endpoint returns the native ZIP as before; with it, the endpoint returns an OpenAPI 3.0 document in YAML (HTTP Content-Type: application/yaml), downloaded as openapi-<timestamp>.yaml.

The same document is also available as JSON by using ?format=openapi-json instead. This returns an identical OpenAPI 3.0 document serialised as JSON (HTTP Content-Type: application/json), downloaded as openapi-<timestamp>.json. Only the serialisation differs; the paths, operations, examples and coverage report described below are the same in both formats. The GUI Export as OpenAPI (YAML) button always downloads YAML; the JSON variant is available through the endpoint.

# Export the active scenario's HTTP mappings as an OpenAPI 3.0 YAML document
curl -X GET "http://localhost:8080/http/management/exportMappings?format=openapi" \
  -o openapi.yaml

# The same document as OpenAPI 3.0 JSON
curl -X GET "http://localhost:8080/http/management/exportMappings?format=openapi-json" \
  -o openapi.json

# Without the format parameter, the same endpoint returns the native WireMock ZIP (unchanged)
curl -X GET "http://localhost:8080/http/management/exportMappings" \
  -o mappings.zip

What gets exported

Traffic Parrot translates the currently active scenario's HTTP mappings into OpenAPI paths and operations:

• Paths and operations: mappings are grouped by URL path and HTTP method. Two mappings on the same path with different methods become one path with two operations.
• URL matchers: an exact URL or path becomes a literal OpenAPI path; a path template such as /users/{id} becomes a templated path with path parameters. A regular-expression URL matcher (urlMatching / urlPathMatching) is converted to a best-effort path template: each segment containing a regex metacharacter is replaced with a synthetic path parameter ({p0}, {p1}, …) so the path is always valid OpenAPI, and the original regex is preserved in the operation's description, marked as an approximation.
• Request matchers: query and header matchers become OpenAPI parameters (in: query / in: header).
• Responses: each mapping's response body is emitted as a literal example under responses[status].content[contentType].example. The content type is taken from the response's Content-Type header, falling back to application/json when the body is JSON and otherwise text/plain.
• Response schema: when the response body is JSON, an inferred JSON schema is emitted alongside the example under responses[status].content[contentType].schema. This is what client generators, SDK codegen and documentation portals consume to produce real models, rather than just a sample value. See Inferred response schema below for how the schema is derived and its caveats.

Inferred response schema

For each JSON response body, Traffic Parrot infers a JSON schema and emits it alongside the retained example. The example shows a concrete sample value; the schema describes the shape of the response, which is what client generators, SDK codegen and documentation portals actually read. Inference is conservative and based on the single example body, so it asserts only what that one body proves:

• The structure is followed recursively: nested objects and arrays produce nested schemas. An array's items schema is inferred from its first element; a mixed-type array therefore takes the first element's shape.
• Numbers are typed by value: an integral number becomes type: integer, a non-integral number becomes type: number. No numeric format (such as int32/int64) is asserted, because it cannot be known from a single value.
• A JSON null becomes nullable: true with no type (the type is unknowable from null alone).
• Every observed object key becomes a property, but no required array is emitted: one example does not prove a field is always present.
• No enum is inferred: a single observed value is not an enumeration of the allowed values.
• An empty array yields items: {} (the element type is unknown but OpenAPI still requires an items schema).

A response body that is not JSON (XML, plain text, or an empty body) gets no schema. The literal example is still emitted, and the mapping is recorded in the coverage report with reason no-schema-inferred (disposition APPROXIMATED), so the lossiness stays explicit.

Coverage report

Some mappings have no clean OpenAPI representation: for example regular expression, JSONPath or XPath body matchers, proxy responses, and mappings whose HTTP method has no OpenAPI equivalent. These mappings are never silently dropped. Every mapping's disposition is recorded in a coverage report so you always know exactly what was and was not faithfully captured. Each mapping is reported as one of:

• REPRESENTED: translated faithfully.
• APPROXIMATED: translated with a documented caveat (for example a regex URL matcher converted to a synthetic path template, or a JSONPath/XPath/regex request-body matcher that cannot be expressed exactly).
• OMITTED: left out of the paths (for example a proxy response; a mapping whose HTTP method has no OpenAPI equivalent: OpenAPI operations are per-verb, so an ANY or GET_OR_HEAD method cannot be expressed as a single operation; or a second mapping that would duplicate an already-exported path and method).

In particular, a “match any method” stub (a stub that does not specify a method, which defaults to ANY) is reported as OMITTED with reason unsupported-method:ANY rather than appearing in paths. It still shows up in the coverage report so you can see it was not captured.

The coverage report is surfaced in the exported document in two ways:

• A root-level OpenAPI extension x-trafficparrot-export-coverage listing every mapping as {mapping, reason, disposition}.
• A one-line summary in info.description, for example 2 mappings exported (1 approximated), 1 omitted.

openapi: 3.0.1
info:
  title: Traffic Parrot exported HTTP mappings
  description: "2 mappings exported (1 approximated), 1 omitted."
  version: 1.0.0
paths:
  /users/{id}:
    get:
      parameters:
        - name: id
          in: path
          required: true
          schema:
            type: string
      responses:
        "200":
          description: Response
          content:
            application/json:
              schema:
                type: object
                properties:
                  id:
                    type: integer
                  name:
                    type: string
              example: '{"id": 1, "name": "Alice"}'
x-trafficparrot-export-coverage:
  - mapping: /users/{id}
    reason: represented
    disposition: REPRESENTED
  - mapping: /search
    reason: regex-url
    disposition: APPROXIMATED
  - mapping: /health
    reason: unsupported-method:ANY
    disposition: OMITTED

Current limitations

The following are not yet supported and are planned for later releases:

• Single example per response: the schema is inferred from one example body; there is no union or merge across multiple examples.
• Single scenario: only the currently active scenario is exported; there is no multi-scenario merge.
• No security schemes are emitted.
• OpenAPI 3.0 only (not 3.1).

Traffic Parrot can export the HTTP traffic it has served as a HAR 1.2 (HTTP Archive) document, so you can drive a load or resilience test through your own engine. Unlike the native ZIP and the OpenAPI export above, both of which export the deduplicated mappings, the HAR export is taken from the live request journal: the ordered, timed stream of inbound requests Traffic Parrot has served during the current session. That preserves the request order and per-request timing a load profile needs, which the deduplicated mappings discard.

On the Export page (HTTP in the top navigation bar, then Export), alongside the Download Mappings and Export as OpenAPI (YAML) buttons, there is a Download recorded traffic (HAR) button. Click it to download the current request journal as a HAR 1.2 file, named recorded-traffic-<timestamp>.har.

The HAR export is served off the same export endpoint as the native ZIP by adding the ?format=har query parameter (the same way ?format=openapi selects the OpenAPI export), so you can also trigger it programmatically. Without the parameter the endpoint returns the native ZIP as before; with it, the endpoint returns a HAR 1.2 document (HTTP Content-Type: application/json), downloaded as recorded-traffic-<timestamp>.har.

# Export the recorded HTTP request journal as a HAR 1.2 document
curl -X GET "http://localhost:8080/http/management/exportMappings?format=har" \
  -o recorded-traffic.har

# Without the format parameter, the same endpoint returns the native WireMock ZIP (unchanged)
curl -X GET "http://localhost:8080/http/management/exportMappings" \
  -o mappings.zip

What gets exported

Each request in the journal becomes one HAR log.entries entry, ordered by the time the request arrived. Every entry carries:

• Request: the HTTP method, the full URL, the request headers, the query string, and the request body (as postData) when one was sent.
• Response: the HTTP status, the response headers, and the response body.
• Timing: startedDateTime (when the request arrived) and time (how long Traffic Parrot took to serve it, in milliseconds), so the relative arrival times and per-request durations are preserved for the load profile.

Driving a load or resilience test

Traffic Parrot does not run the load itself; the HAR file is the hand-off to your own load engine. The exported document is a standard HAR 1.2 file, so it can be fed into the common HAR-to-load-test converters, for example the Gatling HAR Converter (which generates a Gatling simulation from a HAR file) or Grafana har-to-k6 (which generates a k6 script from a HAR file). You then run the generated test on that engine against whatever target you choose.

Journal size limit

The HAR export reads the in-memory request journal, which keeps only the most recent requests, bounded by the trafficparrot.virtualservice.maxRequestJournalEntries property (default 1000; older entries age out as new requests arrive). On a long recording session this means only the most recent 1000 requests are exported. If you need a longer history to export, raise maxRequestJournalEntries in your trafficparrot.properties before recording (a larger journal uses more memory; see the property reference). If the journal is disabled entirely (maxRequestJournalEntries=0, as some performance-tuning profiles set it), the export is a valid but empty HAR with no entries.

First-slice limitations

This is the first release of the HAR export. The following are not yet supported and are planned for later releases:

• Text bodies: request and response bodies are exported as text (the target is HTTP/JSON traffic); binary or Base64-encoded body fidelity is not yet preserved.
• HTTP only: the recorded request journal exported is the HTTP one; there is no messaging or gRPC HAR export.

First, go to HTTP in the top navigation bar and then click Import.

Click the button and select the file(s) to import using the file picker. Traffic Parrot will display the mappings that were successfully imported in the table below.

You can also import programmatically using the Traffic Parrot Management APIs.

• WireMock mappings in ZIP format
• Swagger 1.x and 2.x (both YAML and JSON)
• OpenAPI 2.x and 3.x (both YAML and JSON)
• RAML 0.8
• HAR 1.1 and 1.2 (HTTP Archive files exported from browser DevTools)
• Pact v2, v3 and v4 (consumer-driven contract files, HTTP interactions only)

You can also import directly from a URL; see Importing from a URL.

HAR (HTTP Archive) files let you create stubs from captured network traffic rather than writing them from scratch. To obtain a HAR file:

1. Open the browser DevTools (press F12 or right-click and select "Inspect")
2. Go to the Network tab
3. Interact with the application to capture the HTTP traffic you want to mock
4. Right-click on any request in the network list and select Save all as HAR with content (Chrome) or Save All As HAR (Firefox)

Then import the .har file. Traffic Parrot creates a stub mapping for each HTTP request/response pair, named with a [HAR Import] prefix followed by the status code, HTTP method, and URL path for easy identification.

What gets imported

• Request matching: HTTP method and exact URL path including query string parameters
• Content-Type matching: when the original request included a Content-Type header, the stub also matches on it
• Response: status code, body (base64-encoded bodies are decoded automatically), and Content-Type header

By default, all other request headers (Accept, User-Agent, Cookie, Authorization, etc.) and all non-Content-Type response headers are excluded, to keep stubs flexible and reusable. If you need the stub to reproduce the recorded exchange faithfully (keeping every header), use the Preserve full fidelity (all headers) option described below. HAR files can also be placed inside ZIP archives in the http-import/ directory.

Preserve full fidelity (all headers)

On the import preview screen, next to the Import Selected button, is a Preserve full fidelity (all headers) checkbox. It is off by default, which produces the flexible stubs described above. The checkbox applies to the whole import; it is a bulk setting for all selected entries, not a per-entry one.

When you tick it, every imported stub reproduces its recorded request and response exactly:

• Request: every recorded request header, including Content-Type, becomes an exact request matcher (matched with the full recorded value).
• Response: every recorded response header is emitted on the stub response, including the exact Content-Type with its charset.

The request and response bodies are imported the same way in both modes (base64-encoded bodies are decoded automatically); only header handling changes.

Caution: full fidelity makes stubs strict, and often single-use. Turning every recorded header into an exact request matcher means the stub only matches a request that is byte-for-byte and environment-identical to the recording. Headers such as Host, Cookie and Authorization tie the stub to a specific host, session or token, and a Content-Length matcher only matches a request whose body is exactly the same length. Leave the option off (the default) for reusable stubs, and turn it on only when you specifically want a faithful reproduction of the exact recorded exchange.

HTTP/2 pseudo-headers (:method, :authority, :scheme, :path and :status), which appear in HAR files exported from browsers, are automatically excluded in full-fidelity mode: they are never turned into request matchers or emitted as response headers.

This option applies to HAR import only; it does not affect OpenAPI, Swagger, RAML, Pact or WireMock ZIP imports.

Pact is a consumer-driven contract format. A Pact file records the requests a consumer makes and the responses it expects from a provider, so it is a convenient source of stubs when you want a virtual service to stand in for that provider. Traffic Parrot imports Pact specification v2, v3 and v4 files (the Synchronous/HTTP interactions of a v4 contract).

There is nothing extra to configure. Import the Pact .json file the same way you import any other format: Traffic Parrot detects that the file is a Pact contract automatically, so there is no format to choose. Each HTTP interaction in the contract becomes one stub mapping, named with a [Pact Import] prefix followed by the status code, HTTP method, and URL path. When an interaction has a provider state, the state is appended to the name to make it easy to identify.

What gets imported

• Request matching: HTTP method, URL path and query string, request headers, and request body
• Response: status code, response headers, and response body

JSON request bodies are matched tolerantly (extra elements and array ordering are ignored), which mirrors the way Pact matches a body by structure rather than byte-for-byte. XML and other bodies are matched against the example value from the contract.

The only structural difference between the versions is how the body is stored. In a v4 contract each request and response body is wrapped in a body object that carries the payload under content and its media type under contentType; Traffic Parrot reads the payload from content and uses the declared contentType to decide how to match the body. Everything else (method, path, query, headers, matching rules and provider states) is imported the same way as for v2 and v3.

Pact matching rules

Where a Pact contract declares matchingRules, Traffic Parrot honours the rules that map onto its own request matchers and falls back to an exact match for the rest. When a rule falls back, the import reports a warning listing the affected rules so you know which matchers became literal and may want to relax by hand on the imported mapping.

Message (asynchronous) interactions

This HTTP import handles the HTTP interactions of a Pact contract. A Pact contract can also describe asynchronous message interactions: a top-level messages[] array in a specification v3 contract, or Asynchronous/Messages interactions[] in a v4 contract. Import those as messaging stubs from the messaging pages instead: Import a message Pact contract on the JMS page, or Import a message Pact contract on the native IBM® MQ page. When a v4 contract mixes HTTP and message interactions, this HTTP import brings in the HTTP interactions and skips the message interactions with a warning; a contract that contains only message interactions is reported as having no HTTP interactions to import; use the JMS or IBM® MQ message Pact import for it.

Not supported in this release

This first release imports HTTP interactions only. The following are not yet supported:

• Generating dynamic (templated) responses from a contract: the example responses from the contract are imported as-is.
• Provider verification and pulling contracts from a Pact Broker. Import a contract file you already have on disk.

Instead of downloading a specification to disk and then uploading it, you can paste a URL into the Import page and let Traffic Parrot fetch the document for you. This is convenient when the specification is published at a stable address: for example a running service's /v3/api-docs endpoint, or a hosted OpenAPI, HAR or Pact file.

Traffic Parrot fetches the document server-side and feeds it into the same content-based import pipeline as an uploaded file. The format is detected from the content, so Swagger/OpenAPI, HAR and Pact specifications can be imported by URL exactly the way you would import them from a file. As with file upload, OpenAPI, HAR and Pact documents show a preview before importing.

Enabling the feature

Importing from a URL is off by default and must be enabled by an administrator. Until it is enabled the URL field is not shown on the Import page and the underlying server endpoint stays inert. Set the following two properties in trafficparrot.properties and restart Traffic Parrot:

• trafficparrot.http.importfromurl.enabled=true: turns the feature on. While this is false (the default), the URL field is absent from the Import page.
• trafficparrot.http.importfromurl.allowedHosts: a comma-separated allow-list of host or host:port entries that Traffic Parrot is permitted to fetch from. It is default-deny: when the feature is enabled but the allow-list is empty, every fetch is refused (fail-closed) with a message asking you to configure the allow-list. Only URLs whose host matches an allow-listed entry are fetched.

Both property names are described in full on the Configuration reference page.

The same two properties govern the fetch of an OpenAPI example that points at a URL with externalValue, whether the specification is imported on the Import page (from a file or from a URL), read from the OpenAPI directory at startup or built by the MCP generate_mappings_from_openapi tool. While the feature is off, the example's host is not allow-listed, or that host resolves to one of the blocked addresses listed below, that example is replaced with a Traffic Parrot could not fetch externalValue: <url> (<reason>) placeholder and the mapping is still created.

Why the allow-list is required

Because Traffic Parrot makes the fetch from its own host, an unrestricted version could be used to probe addresses inside your network, a server-side request forgery (SSRF) risk. The host allow-list keeps the feature confined to a set of addresses you trust. In addition, Traffic Parrot rejects URLs that resolve to loopback, link-local, cloud-metadata (169.254.169.254) or private (RFC 1918 / IPv6 unique-local) addresses (even if such a host is allow-listed) guards against DNS rebinding, restricts the scheme to http and https, and caps the response size and the connect and read timeouts. Keep the allow-list as tight as possible.

How to import from a URL

1. Enable the two properties above and restart Traffic Parrot.
2. Go to HTTP in the top navigation bar and click Import.
3. Paste the specification URL into the URL field.
4. Click Fetch & preview.
5. For an OpenAPI, HAR or Pact specification, review and filter the preview and select the entries you want.
6. Click Import Selected to create the stubs.

RAML and WireMock ZIP specifications cannot yet be imported from a URL; for those formats, download the file and import it from disk instead.

When you upload a HAR, OpenAPI, or Pact file, Traffic Parrot shows a preview of the entries before importing. This lets you review, filter, and select exactly which stubs to create.

Filtering entries

The preview panel provides four filters that you can use to narrow down the entries:

• Host: dropdown auto-populated with the unique hosts found in the file
• Method: dropdown auto-populated with the HTTP methods found in the file (GET, POST, PUT, DELETE, etc.)
• Status: dropdown to filter by status code range (2xx, 3xx, 4xx, 5xx)
• URL contains: text input to filter by URL. It matches against the request path (and query string), not the host (use the Host dropdown for that). Plain text matches any path that contains it (e.g. users matches /api/users). If the text includes a glob wildcard (* for any sequence of characters or ? for any single character), the filter switches to an anchored, case-insensitive glob match where the whole path must match the pattern. For example /api/* matches paths beginning /api/, *.json matches paths ending .json, and /users/? matches /users/1 but not /users/12. To match a literal ? in a path, filter by plain substring without any other wildcard.

All filtering happens instantly in the browser without extra server requests. Filters combine with AND logic (e.g. selecting host "api.example.com" and method "POST" shows only POST requests to api.example.com).

This is a preview convenience that controls which entries are shown for selection; it does not change how the imported stubs match requests at runtime.

Selecting entries

Each entry has a checkbox. Use the header checkbox to select or deselect all visible (filtered) entries at once. You can also toggle individual entries. The "Import Selected" button shows the count of selected entries and is disabled when no entries are selected.

Duplicate detection and conflict resolution

The Duplicate column shows whether each entry is new or already exists. Entries that do not match any existing stub display a green "New" badge and are imported as-is. If an entry matches an existing stub with the same HTTP method and exact request URL (one created from a different import or added by hand), a yellow "Duplicate" badge appears instead.

Previously a duplicate could only be skipped. Now each duplicate entry offers an explicit action in the On duplicate column, and a single For duplicates: control above the table sets the action for all duplicates in one click. A per-row choice overrides the bulk default for that row. The three actions are:

• Skip (the default): do not import the duplicate; the existing stub is left unchanged. This is the behaviour of earlier releases, so imports you do not touch keep working exactly as before.
• Overwrite: remove the existing stub for that method and URL and add the imported one, so no duplicate is left behind. If more than one existing stub matched, all of them are replaced by the single imported stub.
• Import as new: keep both. The existing stub is untouched and the imported stub is added under a disambiguated name (<name> (copy), then (copy 2), (copy 3), and so on) so the two are easy to tell apart in the mappings list.

Re-importing the same file still replaces the mappings it created last time, so those entries show as New rather than Duplicate; the conflict actions apply to a clash against a stub from a different import or one you created manually. To keep two revisions of the same endpoint side by side, capture them under different file names and choose Import as new.

Known limitation: duplicate detection compares the HTTP method and the exact request URL string. An existing stub that matches its URL with a pattern or regular expression (rather than an exact URL) is not recognised as a duplicate, so the imported entry is added as a separate stub whichever action you choose.

Importing selected entries

Click "Import Selected" to create stubs only for the checked entries. The imported mappings appear in the table below. Click "Cancel" to discard the preview and return to the file selector.

For HAR imports, a Preserve full fidelity (all headers) checkbox sits next to the "Import Selected" button. Leave it unticked for flexible stubs, or tick it to reproduce every recorded request and response header exactly; see Preserve full fidelity (all headers).

Format support

The preview and filter flow is available for HAR, OpenAPI/Swagger, and Pact files. RAML files and WireMock ZIP archives are imported directly without a preview step.

You can import mapping files programmatically using the REST API instead of the web UI. This is useful for CI/CD pipelines and automation scripts.

Send a POST request to /http/management/importMappings with the file(s) as a multipart form upload:

# Import a single file (HAR, OpenAPI/Swagger, RAML, or WireMock ZIP)
curl -X POST http://localhost:8080/http/management/importMappings \
  -F "files[]=@capture.har"

# Import multiple files at once
curl -X POST http://localhost:8080/http/management/importMappings \
  -F "files[]=@api-spec.yaml" \
  -F "files[]=@capture.har"

On success, the response contains the IDs of the imported mappings:

{"mappings":["id1","id2",...]}

When part of an upload could not be imported but at least one mapping was, the response also carries a warnings array with one message per file, or part of a file, that could not be converted into a mapping. An upload from which nothing could be imported is refused with HTTP 400 instead.

{"mappings":["id1"],"warnings":["Did not find any mappings to import in other-file.yaml!"]}

Each imported mapping is registered under a fresh id and records the name of the uploaded file it came from as tp.import.source in its metadata. Importing a file of the same name again first removes the mappings its previous import registered, so a re-import replaces the earlier import rather than adding a second copy of each mapping. A mapping registered any other way is not removed by an import, but an imported mapping is written to the file its name maps to: a mapping whose name ends in .json is written to the file of that name, so importing an instance's own export writes each such mapping over the file of the mapping it was exported from, and the imported mapping takes its place.

On failure, the response returns HTTP 400 with an error message:

{"result":"error message"}

An upload larger than trafficparrot.multipart.upload.max.size.mb (default 10 MB) is rejected with HTTP 413 and a {"result": …} error body, and the mappings are not imported.

This endpoint supports all supported import formats: HAR, Swagger/OpenAPI, RAML, Pact, and WireMock ZIP.

Preview before importing

You can preview the entries that would be created before committing to an import. Send a POST request to /http/management/previewImport:

curl -X POST http://localhost:8080/http/management/previewImport \
  -F "file=@capture.har"

The response contains a list of entries with their method, host, path, and status code:

{
  "previewSupported": true,
  "entries": [
    {"index": 0, "method": "GET", "host": "api.example.com", "pathWithQuery": "/users", "statusCode": 200, "duplicate": false},
    {"index": 1, "method": "POST", "host": "api.example.com", "pathWithQuery": "/users", "statusCode": 201, "duplicate": false}
  ],
  "hosts": ["api.example.com"],
  "methods": ["GET", "POST"]
}

For file types that do not support preview (RAML, WireMock ZIP), the response returns {"previewSupported": false}.

Filtered import

To import only specific entries, send a POST request to /http/management/importSelectedEntries with the file and filter parameters:

# Import specific entries by index (from the preview response)
curl -X POST http://localhost:8080/http/management/importSelectedEntries \
  -F "file=@capture.har" \
  -F "selectedIndices=0,2,5"

# Import entries matching filter criteria
curl -X POST http://localhost:8080/http/management/importSelectedEntries \
  -F "file=@capture.har" \
  -F "hostFilter=api.example.com,auth.example.com" \
  -F "methodFilter=GET,POST" \
  -F "statusFilter=2xx,4xx" \
  -F "urlFilter=/api/v1,/users"

Available filter parameters (all accept comma-separated values):

Filters combine with AND logic across filter types and OR logic within each filter type. For example, hostFilter=api.example.com&methodFilter=POST,PUT imports POST and PUT requests to api.example.com. If no filter parameters or indices are provided, all entries are imported.

Preserving full fidelity (all headers)

Both /http/management/importMappings and /http/management/importSelectedEntries accept an optional preserveAllHeaders parameter, the REST equivalent of the Preserve full fidelity (all headers) checkbox in the UI. Set it to true to turn every recorded request header into an exact request matcher and emit every recorded response header on the stub response. Omit it (or set it to false) for the default flexible behaviour.

# Direct upload, preserving all headers
curl -X POST http://localhost:8080/http/management/importMappings \
  -F "files[]=@capture.har" \
  -F "preserveAllHeaders=true"

# Filtered import, preserving all headers
curl -X POST http://localhost:8080/http/management/importSelectedEntries \
  -F "file=@capture.har" \
  -F "selectedIndices=0,2,5" \
  -F "preserveAllHeaders=true"

Resolving duplicate entries

When a selected entry duplicates an existing stub (same method and exact URL), you can control what happens to it, the REST equivalent of the conflict-resolution dropdowns in the UI. Pass an optional conflict-strategy parameter to set the action for every duplicate in the request, and an optional conflictStrategies parameter to override it for individual entries by index:

# Overwrite every duplicate with the imported entry
curl -X POST http://localhost:8080/http/management/importSelectedEntries \
  -F "file=@capture.har" \
  -F "selectedIndices=0,1,2" \
  -F "conflict-strategy=overwrite"

# Skip duplicates by default, but overwrite entry 0 and import entry 2 as a new stub
curl -X POST http://localhost:8080/http/management/importSelectedEntries \
  -F "file=@capture.har" \
  -F "selectedIndices=0,1,2" \
  -F "conflictStrategies=0:overwrite,2:importAsNew"

This parameter affects HAR imports only; it is ignored for other formats. Full fidelity makes stubs strict, so review the caution above before enabling it for reusable stubs. HTTP/2 pseudo-headers are excluded automatically.

overwrite removes every existing stub that matches the entry's method and URL before adding the imported one; importAsNew keeps the existing stub and adds the imported one under a disambiguated name (<name> (copy)). As in the UI, duplicate detection is an exact method-and-URL match, so an existing stub that matches its URL by pattern or regular expression is not treated as a duplicate.

When importing from a specification format such as Swagger or OpenAPI, it is possible to use dynamic responses in the example responses in the specification.

Dynamic responses typically result in usage of the {{...}} notation that is not valid JSON or YAML. In these cases, you will need to specify the response as an escaped string when using JSON or YAML based import formats such as Swagger and OpenAPI.

You can find examples of dynamic OpenAPI responses in the examples project.

Traffic Parrot also supports selecting the response to return using a special test request header

x-traffic-parrot-select-response-status

that contains the status code of the response to return.

In this way, you can define OpenAPI examples for each status code, and then when you import them into Traffic Parrot, you can select a response by providing the test header with the response status code.

POST /example HTTP/1.1
x-traffic-parrot-select-response-status: 400

HTTP/1.1 400 Bad Request
{
  "example": "Message from OpenAPI example"
}

trafficparrot.openapi.import.mode=SELECT_RESPONSE_STATUS
trafficparrot.openapi.skeletons.mode=SELECT_RESPONSE_STATUS

• When OpenAPI response examples are present, they are used directly as mock responses
• Otherwise, OpenAPI schema data types and field structures are used to generate a valid mock response
• The request header x-traffic-parrot-select-response-status can be set to a numeric response code to select which response to return
• The default response returned is the success response
• The request body is not included in matching, to simplify the response selection
• The request URL is used for matching, including checking for mandatory path and query parameters

First, go to HTTP in the top navigation bar and then click Add/Edit.

Fill in the Request/Response fields and click Save to configure a mapping.

You can also select from the HTTP skeleton dropdown which will populate the Request/Response fields for that skeleton.

After saving the mapping, it will appear in the list of mappings.

Clicking the edit button will allow you to edit an existing mapping.

The Edit Mapping panel has a dedicated Query parameters panel below the Request / Response panels that lets you match individual query string parameters without putting them in the URL field or hand-writing a URL regex.

When a mapping has no query parameter matchers, the panel shows only its heading and the Add query parameter button. Click Add query parameter to add the first row; each additional click appends another row. Each row exposes a parameter name, a matcher dropdown, an optional value, and a remove button.

Saving the mapping produces standard WireMock queryParameters JSON. Each row becomes one entry, keyed by parameter name:

{
  "request": {
    "method": "GET",
    "urlPath": "/search",
    "queryParameters": {
      "q":       { "equalTo": "test" },
      "page":    { "matches": "[0-9]+" },
      "preview": { "absent": true }
    }
  },
  "response": {
    "status": 200,
    "body": "..."
  }
}

Loading a mapping that already contains a queryParameters block populates the rows on the form; the round-trip is preserved, so editing and re-saving a mapping does not change or drop existing query parameter matchers.

A mapping with queryParameters matches loosely: a request that includes additional query parameters not listed in the mapping still matches, provided every listed parameter satisfies its matcher. To match only on the path, use urlPath or urlPathPattern in the URL field; using url or urlPattern together with queryParameters is supported but the URL must not itself include a query string.

Caution: when matching on query parameters, set the URL match mode to path equal to or path matches regex, not equal to or matches regex, which match the whole URL (including the query string) and will conflict with the separate query parameter matchers. The Edit Mapping panel shows an inline warning beneath the URL match mode if you combine a full-URL mode with one or more query parameter rows.

When to use Query parameters vs. the URL field

The older approaches (full query string in urlEqualTo, lookahead regex in urlMatching) continue to work for existing mappings; the new section is additive. For most cases though, the Query parameters section produces a mapping that is easier to read, easier to edit, and tolerant of parameter ordering in the incoming request.

Pasting a full URL (e.g. one copied from a browser or a curl command) into the Request URL field leaves the query string baked into the URL value, which means the mapping then matches the exact query string only. To convert that query string into per-parameter rows in the Query parameters section in one click, use the Convert query string to rows link that appears directly below the Request URL input on the Edit Mapping panel.

Clicking the link parses each name=value pair from the URL's query string, appends one row per pair to the Query parameters section with the matcher set to equal to, and strips the query string from the Request URL field. URL-encoded characters in values (e.g. %20) are decoded automatically. The link does not save the mapping; the form changes are staged in the browser and you still need to click Save to commit them.

Before clicking the link, the Request URL field still contains the full URL with its query string and the Query parameters section is empty:

After clicking the link, the query string has been stripped from the URL and one equal to row has been appended for each name=value pair:

Example: with the Request URL set to /search?q=hello%20world&page=1, clicking Convert query string to rows rewrites the URL field to /search and adds two rows under Query parameters: name q with matcher equal to and value hello world, and name page with matcher equal to and value 1. Existing rows are not touched; converted rows are appended to whatever rows are already present.

If the URL has no query string the link is a no-op. Multi-occurrence parameters (e.g. ?tag=a&tag=b) produce one row per occurrence, but because the saved queryParameters JSON is keyed by parameter name, only the last row for a given name takes effect at match time. To match the same parameter appearing more than once, keep using the urlMatching lookahead workaround described in the Query parameters section above.

The request priority can be set in order to set up a preference order for matching mappings. This works in the same way that priority works in WireMock.

The highest priority value is 1. If two or more mappings both match a request, the mapping with the higher priority will be used to provide the response. The default priority is 5.

This can be useful, if you want a "catch-all" mapping that returns a general response for most requests and specific mappings on top that return more specific responses.

You can temporarily disable a mapping without deleting it. Disabled mappings remain in the mapping list but are skipped during request matching, so no incoming request will match them.

In the mapping list, each mapping has a toggle button (eye icon) in the actions column. Click the toggle to disable a mapping. The mapping row appears dimmed to indicate it is disabled. Click the toggle again to re-enable it.

This is useful for:

• Temporarily removing a mapping from matching without losing its configuration
• Simulating service failures by disabling specific response mappings
• A/B testing different response configurations

The enable/disable state is stored in the mapping's metadata as tp.enabled. Mappings without this metadata field default to enabled, so existing mappings are not affected.

Traffic Parrot supports responding with a HTTP response including headers, body and status code, from another HTTP server.

This can be used to forward the requests you don't have mappings for to another endpoint.

You can use this to provide a default response for unmatched requests:

1. Set a low priority like 10 for the mapping with the proxy response
2. Leave default priority for other non-proxy response mappings
3. Now any unmatched requests will return a response from the proxy mapping

Alternatively, you can configure automatic passthrough for all unmatched requests using the HTTP Passthrough Proxy property. This creates the catch-all proxy mapping automatically on startup.

Traffic Parrot can also be configured to send an additional HTTP request after the mock response has been returned. This is called HTTP webhooks or callbacks.

It is possible to enable the callback based on a script evaluation condition. For example, this can be used to only send the callback if the mock response contains a particular success code.

The callback URL, headers and body are Handlebars templates, so a callback can carry the signature a real provider would send, computed over the callback body with the {{hmac}} helper. See Signing an outbound webhook for a complete example.

Another way to match HTTP requests is with a request matcher script. If the expression evaluates to true, then the request is matched, otherwise it is not matched.

A script can also verify a signature on the incoming request, so that the mock accepts only correctly signed callbacks from the system under test; see Verify an inbound webhook signature.

HTTP skeletons offer a quick way to generate a template of a mapping that matches an endpoint defined in an OpenAPI specification file. We may also add support for other specifications e.g. RAML in the future.

In order to edit the list of elements on the HTTP skeletons dropdown on the Add/Edit page you will need to place OpenAPI specifications into the file trafficparrot-x.y.z/openapi configuration directory. JSON and YAML OpenAPI files are supported.

Alternatively, you can use the button to import OpenAPI files. If you import a file with the same name as a file that was previously imported, it will be overwritten.

Real services are slow before they are broken, and they break in more ways than an error status code. Traffic Parrot can hold a response back for a fixed or randomised time, dribble the body out over several seconds, or inject a connection-level fault such as a reset, an empty response or a malformed chunk. That lets you exercise the timeout, retry, circuit-breaker and error-handling paths of the system under test on demand, instead of waiting for the real service to misbehave.

Delays and faults are set per mapping, on the Response panel of the Add/Edit mapping form. There is no server-wide switch and no property to set: a mapping that carries a delay or a fault behaves that way every time it matches, and every other mapping is unaffected.

Coming from WireMock? The fixedDelayMilliseconds, delayDistribution, chunkedDribbleDelay and fault response attributes work unchanged, so mappings you already have (including ones you import as a WireMock ZIP) load and serve without modification. Traffic Parrot adds five faults of its own on top of the four standard ones.

The Successful tab of the Response panel carries the delay controls. Fill in the one that matches the shape of slowness you want to reproduce, then save the mapping as usual.

Fixed delay. Type a number of milliseconds into the first Response delay field and leave the second one blank. Every request that matches the mapping waits that long before the response is written.

Uniform random delay. Fill in both Response delay fields. Each response waits a random number of milliseconds drawn evenly from the range, with both bounds included, which is the way to add jitter around an average. Two identical values are saved as a fixed delay instead, because there is nothing to randomise.

Log-normal random delay. Real response times rarely cluster evenly: most calls land near a typical figure and a minority take much longer. A log-normal delay reproduces that long-tailed shape. Set ms median to the 50th percentile you want in milliseconds, sigma to the standard deviation of the underlying normal distribution (unitless, and the larger it is the longer the tail), and optionally ms max to truncate the tail so no single response can exceed it. A maximum below the median is rejected, so set it above the median or leave it blank. As the note under the fields says, when median and sigma are both set the log-normal delay overrides the fixed or uniform delay above it.

A delayed response holds a virtual service worker thread for the whole of the delay, so a long delay combined with high concurrency can exhaust the thread pool; that is worth keeping in mind for load and soak runs rather than for a handful of functional tests.

A delay makes the caller wait and then hands over the whole body at once. A chunked dribble delay does the opposite: the response starts arriving immediately but trickles in, which is what a saturated or bandwidth-starved service looks like from the client side.

Fill in the two Chunked dribble delay fields on the Successful tab: the number of chunks the response body is split into, and the total duration in milliseconds over which those chunks are sent. Five chunks over 2000 ms sends roughly a fifth of the body every 400 ms, so the caller waits about two seconds for the last byte.

The caller receives the complete body and a well-formed response, just slowly, so this exercises read timeouts, progress handling and streaming parsers rather than an error path. It is saved on the mapping as:

"chunkedDribbleDelay": {
  "numberOfChunks": 5,
  "totalDuration": 2000
}

A fault damages the response at the HTTP or connection level rather than returning a different status code, so the caller has to cope with a failure its HTTP client cannot turn into a normal response object. Traffic Parrot offers nine of them, in two places on the Response panel.

The Response fault dropdown on the Successful tab holds the faults that damage a response you have otherwise configured. The status code, headers and body you set are still sent; the fault changes how the body reaches the client. It offers Close connection after half of response, Dripping bytes (100ms each), Slow dripping bytes (1s each) and Very slow dripping bytes (10s each).

The Faulty tab's Response dropdown holds the faults that replace the response altogether. Picking one here saves the fault as the entire response definition, so the status code, headers, body and delays on the Successful tab are not saved with it. It offers Empty response, Malformed response chunk, Random data and close, Incorrect too small Content-length and Connection reset by peer.

One of those needs a response body to act on: Incorrect too small Content-length under-declares the length of the body it is sending, so a mapping saved from the Faulty tab (which keeps no body) has nothing to truncate. To get the truncated read, give the mapping a body by writing that mapping as JSON, where the fault and the body sit side by side.

The value in the left-hand column is what appears in the mapping's fault attribute. Values are case-sensitive, including the lowercase ms and s suffixes on the dripping faults.

The dripping faults spend one interval per byte, so a response takes its body length multiplied by the interval: a 40-byte body dripped at one byte per second takes about 40 seconds, and the same body at ten seconds a byte takes most of seven minutes. Keep the body short on a dripping mapping, or the test waiting on it will outlast your patience.

Everything the form saves is stored as standard response attributes on the mapping, so you can write the same thing by hand. That is the path to take when mappings are generated by a script or kept in version control, and it is how you combine a fault with a body, headers or a status code that the Faulty tab does not save.

Drop the mapping into the mappings directory as mappings/slow-order.json:

{
  "request": {
    "method": "GET",
    "url": "/orders/1"
  },
  "response": {
    "status": 200,
    "headers": { "Content-Type": "application/json" },
    "body": "{\"id\": 1, \"status\": \"CONFIRMED\"}",
    "delayDistribution": {
      "type": "lognormal",
      "median": 800,
      "sigma": 0.1
    }
  }
}

Or push it into a running instance through the management API, on the management port (trafficparrot.gui.http.port, 8080 by default):

curl -X POST http://localhost:8080/api/http/mappings \
  -H "Content-Type: application/json" \
  -d '{
    "request": { "method": "GET", "url": "/orders/2" },
    "response": {
      "status": 200,
      "body": "The full body whose Content-Length is under-declared",
      "fault": "INCORRECT_SMALL_CONTENT_LENGTH"
    }
  }'

Then call the stub on the virtual service port (trafficparrot.virtualservice.http.port, 8081 by default) and watch it misbehave:

curl -i http://localhost:8081/orders/2

Mappings authored this way open in the Add/Edit form with their delay and fault fields filled in, and saving from the form keeps them, so you can start in JSON and carry on in the UI. The one combination to keep in JSON is a Faulty tab fault paired with a response body, such as the INCORRECT_SMALL_CONTENT_LENGTH example above: the form saves that fault as the whole response, so re-saving the mapping there would drop the body.

The response attributes are:

These attributes are described in the mapping JSON Schema that every instance serves at http://localhost:8080/schemas/http-mapping.schema.json, which is also what you would hand to a code assistant writing mappings for you; see Point the agent at the mapping schema.

Traffic Parrot can validate your HTTP mappings against OpenAPI specification files at startup. This helps you detect drift between your API specifications and your virtual service mappings, for example, when an endpoint is renamed or removed from the specification but the corresponding mapping is not updated.

When enabled, Traffic Parrot checks that each HTTP mapping's URL path and HTTP method match an operation defined in at least one of the loaded OpenAPI specification files. If any mappings reference endpoints not found in the specifications, startup is prevented and detailed error messages are logged.

Enabling the check

To enable the OpenAPI schema check, set the following property in trafficparrot.properties:

trafficparrot.virtualservice.openapi.check.mapping.schema.on.startup=true

By default this property is set to false, so the check does not run and existing behaviour is unchanged.

Setting up OpenAPI specifications

Place your OpenAPI specification files (JSON or YAML, versions 2.x and 3.x) in the scenarios/{scenario}/openapi/ directory. If multiple specification files are present, all operations from all files are aggregated before the check runs.

If no OpenAPI specification files are found, the check passes silently and startup proceeds as normal.

What is checked

• Each HTTP mapping's URL path is compared against the paths defined in the OpenAPI specifications
• The HTTP method of each mapping must match a method defined for that path in the specification
• URL matching supports exact paths, path-only matching, and path templates with parameters; mappings using regex URL patterns are skipped
• Mappings without a method or with the ANY method are skipped

Error reporting

When violations are found, Traffic Parrot prevents startup and logs detailed error messages including the mapping ID, URL, HTTP method, and the reason for the mismatch. This is similar to the gRPC schema check feature.

The trafficparrot validate command checks your mappings against their bound schemas without starting the server, so you can run it as a step in a continuous integration pipeline. It loads the mappings from a files-root directory, validates each one against the OpenAPI specifications (for HTTP) and proto files (for gRPC) in that directory, prints a report, and exits with a status code your pipeline can act on. This is the headless equivalent of the startup-time OpenAPI schema check and gRPC schema check, with an additional under-coverage check (described below) for HTTP mappings, gRPC mappings, and messaging mappings. It also checks your messaging (JMS, native IBM® MQ and file-message) mappings against your imported AsyncAPI specifications; see Messaging coverage check on the JMS page.

The command performs two complementary checks for each protocol, giving you a bidirectional view of how your mappings and your schemas line up. For HTTP the schema is your OpenAPI specifications; for gRPC it is your proto files; for messaging it is your AsyncAPI specifications:

• Schema drift: every mapping must match a declared schema element. For HTTP, a mapping whose URL path and method are not found in any loaded specification is reported as drift (for example, an endpoint that was renamed or removed from the specification but whose mapping was not updated). For gRPC, a mapping referencing a method that is not present in the bound proto files is reported.
• Under-coverage: every declared schema element must have at least one backing mapping. For HTTP, an operation (a path and method) that no mapping covers is reported; for gRPC, a proto method that no mapping covers is reported; for messaging, an AsyncAPI operation whose destination no mapping covers is reported. This means a partially-mocked API or service cannot silently drift from its contract. Operations, proto methods and destinations you have deliberately chosen not to mock can be excluded with an allowlist.

Pass the files-root directory (the directory that contains your mappings/ and openapi/ directories) as the first argument:

trafficparrot validate /path/to/files-root

The OpenAPI specifications are read from <files-root>/openapi/ (JSON or YAML, versions 2.x and 3.x; all files in the directory are aggregated before the checks run) and the HTTP mappings are read from <files-root>/mappings/. If no OpenAPI specifications are present, the HTTP schema-drift and under-coverage checks contribute nothing and only the gRPC check runs.

When everything lines up, the command prints No mapping schema drift detected. and exits 0. When drift or an uncovered operation is found, it prints one labelled section per problem class. For example, a specification declaring four operations with only one of them mapped reports:

Some OpenAPI operations have no backing mapping (under-coverage): (3)
No mapping covers operation: DELETE /pets/{petId}
No mapping covers operation: GET /pets/{petId}
No mapping covers operation: POST /pets

The path shown is the effective path, including any base path declared in the specification's servers entry (for example, GET /api/v1/items for a specification served under /api/v1).

Mappings that use a regular-expression URL pattern, or the ANY HTTP method, cannot be matched to a specification operation and are skipped when deciding whether an operation is covered. An operation reachable only through such a mapping is therefore reported as uncovered.

The same under-coverage check applies to gRPC mappings: every method declared in the bound proto files must have at least one backing gRPC mapping. The gRPC under-coverage check and its grpc-coverage.properties allowlist are documented on the gRPC page; see Validating gRPC mappings.

The same under-coverage check applies to messaging mappings: every operation declared in your imported AsyncAPI specifications (read from <files-root>/asyncapi/) must have at least one backing messaging mapping: JMS, native IBM® MQ or file-message. A declared operation that no mapping covers is reported as under-coverage. The messaging under-coverage check and its messaging-coverage.properties allowlist are documented on the JMS page; see Messaging coverage check.

The command exits with one of three status codes so a pipeline can fail the build on drift while still distinguishing a genuine drift failure from a misconfiguration:

The check fails closed: a files-root with no mappings (a wrong path or an empty mappings/ directory) is reported as a configuration error (2), never as a clean 0.

If your virtual service intentionally mocks only part of an API, you can mark the remaining operations as deliberately unmocked so that they do not fail the under-coverage check. Create a file named openapi-coverage.properties in the files-root directory (next to your openapi/ and mappings/ directories) and list the excluded operations in the exclude.operations property:

exclude.operations=GET /pets/{petId}, POST /orders

Each entry is a METHOD path token. Entries are separated by commas. The method is case-insensitive. The path is the effective path (including any base path from the specification's servers entry), written exactly as it appears in the report (with {parameter} placeholders). If the file is absent, no operations are excluded.

To stop the allowlist from quietly going stale, an entry that does not match any operation in the loaded specifications is itself reported as drift and causes a non-zero exit code. Remove or correct an entry once the operation it refers to no longer exists in the specification.

gRPC has a parallel allowlist, configured through a grpc-coverage.properties file with an exclude.methods property; see Validating gRPC mappings for the gRPC allowlist specifics. Messaging has a parallel allowlist too, configured through a messaging-coverage.properties file with an exclude.destinations property; see Messaging coverage check for the messaging allowlist specifics.

Run the command as a pipeline step and check its exit status. With the default shell behaviour the non-zero exit code on drift fails the step automatically; you can also inspect $? explicitly:

trafficparrot validate /path/to/files-root
if [ "$?" -ne 0 ]; then
    echo "Mappings have drifted from the OpenAPI specification"
    exit 1
fi

Running the check on every change keeps your virtual service mappings in step with the API contract they stand in for: a renamed or removed endpoint, or a new operation that has not yet been mocked, fails the build instead of silently reaching a downstream environment.

Traffic Parrot helps with SOAP and XML mocking by providing an XML editor, the matchesXML request body matcher and dynamic response helpers such as xPath and xPathList.

Please see the video below for a full demo of recording SOAP requests and responses, and then generating dynamic responses.

Please download Traffic Parrot and the sample UV Index SOAP application and follow the demo in the video below.

If you cannot see a video frame below please download it here

To record HTTPS traffic against any host, point your browser or application's proxy setting at Traffic Parrot and import the Traffic Parrot CA certificate into the client's trust store. See the configuration guide Recording HTTPS via the browser proxy.

If your system requires the use of an outbound HTTP proxy (e.g. in a corporate environment) you will need to make sure Traffic Parrot was started using those proxy settings so that it can correctly record. See the configuration guide HTTP proxy settings.

If you need to record a system with a certificate that is not trusted by default, you will need to specify the trust store properties. See the configuration guide Outbound HTTPS certificates.

If you need to record a system that requires a client certificate for authentication, you will need to specify the key store properties. See the configuration guide Outbound HTTPS certificates.

First, navigate to the recording page by clicking HTTP in the top navigation bar and then click Record.

In order to record traffic to a URL, simply enter the Recording URL and click Start recording.

All traffic received by the Traffic Parrot virtual service will be proxied to the host in the Recording URL and recorded as mappings.

To record HTTPS traffic to any host without specifying a single Recording URL, point your client's proxy setting at Traffic Parrot instead. See Recording HTTPS via the browser proxy.

Clicking the edit button will allow you to edit the recorded mapping.

If there are existing mappings present before a recording starts, these mappings will be used to return responses instead of recording a new response.

For example, if there is a mapping with path equal to /example then any traffic to path /example will return a response from the existing mapping and traffic to other paths will still be recorded.

You can delete a recorded mapping while a recording session is still active, for example to prune noise as it is captured. When you delete a mapping during recording, Traffic Parrot remembers that endpoint for the rest of the current recording session, so the same request is not recorded again if it is proxied a second time.

The endpoint is matched by HTTP method and normalised URL path, so query-string variants and numeric-id variants (for example /orders/123 and /orders/456) of the deleted request are suppressed as well.

While a recording session is active, the mappings list shows a Recording active banner above the table as a reminder that deleting a mapping here also stops that endpoint being re-recorded for the rest of the session. The banner is shown only while recording is running.

While a recording session is active, the Record page shows a Live coverage panel that refreshes about once a second. It answers the question “has this recording captured the endpoints I expected, and do they all have mappings yet?” without waiting for the session to finish. The panel is empty when you are not recording (it shows Start recording to see live coverage) and fills in as traffic is proxied through Traffic Parrot.

An endpoint is one HTTP method plus a normalised URL path. Numeric and variable path segments are collapsed using the same convention that Mapping cleanup uses (for example /orders/123 and /orders/456 are counted as the one endpoint /orders/[0-9]+), and query-string variants of the same path are not counted as separate endpoints. This means the coverage tally and cleanup always agree about what “the same endpoint” is.

The summary line above the table reports the totals for the current scenario: how many distinct endpoints have been seen, how many of those now have a mapping, how many were observed but have no mapping yet, and how many mappings exist that have not been observed during this session.

Each row in the table shows one endpoint:

Watch the Observed but unmapped rows: those are endpoints your traffic hit that have not produced a mapping yet (for example a request that was answered by an existing mapping rather than recorded, or one that was filtered out). They are the “you hit this but it didn’t get a mapping” case.

The Live coverage panel reads from an endpoint on the Traffic Parrot GUI/admin port that you can also poll yourself. Send a GET request to /api/http/currentRecordingCoverage:

curl http://localhost:8080/api/http/currentRecordingCoverage

While recording is active, the response carries "recording": true and the per-endpoint coverage for the current scenario:

{
  "recording": true,
  "coverage": {
    "endpoints": [
      {
        "method": "GET",
        "path": "/orders/[0-9]+",
        "mapped": true,
        "mappedCount": 1,
        "observedCount": 3,
        "unmatchedCount": 0
      },
      {
        "method": "POST",
        "path": "/orders",
        "mapped": false,
        "mappedCount": 0,
        "observedCount": 1,
        "unmatchedCount": 1
      }
    ],
    "totalEndpoints": 2,
    "mappedEndpoints": 1,
    "observedButUnmappedEndpoints": 1,
    "mappedButNeverObservedEndpoints": 0,
    "totalRequestsInJournal": 4
  }
}

When no recording session is running, the response is well defined: the flag is false and the coverage totals are zeroed with an empty endpoint list. This is the same idle state the panel renders as Start recording to see live coverage.

{
  "recording": false,
  "coverage": {
    "endpoints": [],
    "totalEndpoints": 0,
    "mappedEndpoints": 0,
    "observedButUnmappedEndpoints": 0,
    "mappedButNeverObservedEndpoints": 0,
    "totalRequestsInJournal": 0
  }
}

If the Recording URL includes a path, only traffic to this path will be recorded, however all traffic will still be proxied.

For example, if the Recording URL is set to http://example.com/aSampleResource then only traffic to paths staring with /aSampleResource will be recorded and all other traffic will be proxied to http://example.com without saving mappings.

By default the Record page shows a single Recording URL input for the common case of recording one backend. To record from multiple backends in a single session, click Record from multiple URLs below the input. This reveals a URL pattern field alongside the target field and adds a second row so you can route different paths to different backends. Use Add proxy target to add more rows, or click the remove button on a row to delete it. Removing rows until only one row remains with an empty pattern returns the form to the simple single-URL mode.

Each row consists of a URL pattern (a regular expression matched against the request path) and a target URL. For example:

The path is used to determine which API should be used to provide the recorded response. The expression .* is used as a wildcard matcher.

• /general/something will route to http://127.0.1.1:8090
• /general/specific/something will route to http://127.0.1.1:8091
• /something will route to http://127.0.1.1:8092

By default, every response body captured during recording is written to a separate file in the __files/ directory and the recorded mapping references it by name. You can instead keep JSON response bodies inline in the mapping itself, so a recorded JSON API produces a single self-contained mapping file with no companion body file.

On the HTTP → Record page, click Advanced parameters and tick Keep valid-JSON response body inline, then start recording as usual.

The checkbox appears below the Record request headers for matching field in the same Advanced parameters accordion.

When the option is on, a recorded response body is kept inline in the mapping (as a structured jsonBody) when both of these are true:

• the response is served with a text content type (for example application/json); and
• the response body is valid JSON.

Any other response body is still written to a __files/ file exactly as before. In particular:

• a non-JSON text body (HTML, XML, plain text, …) is externalised to a file;
• a binary body is externalised to a file: even if its bytes happen to parse as JSON, a body served with a binary content type such as application/octet-stream is written to a file, because the declared content type is honoured.

The inline JSON is stored byte-for-byte as it was received; it is not reformatted. When the option is off (the default), all response bodies are externalised to __files/ exactly as in earlier releases, so existing recording workflows are unchanged.

The checkbox controls a single recording session. To change the default state of the checkbox for everyone (for example to have it ticked out of the box), set the trafficparrot.http.recording.response.inlineJson property in trafficparrot.properties. The property only seeds the checkbox's initial state: whatever the checkbox is set to when you start recording is what takes effect for that session.

By default, when a recorded request has a body, that body is written inline into the recorded mapping as the request body matcher (a bodyPatterns value). You can instead have the request body externalised to a file in the __files/ directory, with the recorded mapping's request matcher carrying a bodyFileName reference to that file instead of the inline value, the request-side counterpart of a file-backed request body matcher.

On the HTTP → Record page, click Advanced parameters and tick Externalise request body to file, then start recording as usual.

The checkbox appears below the Keep valid-JSON response body inline checkbox in the same Advanced parameters accordion.

When the option is on, a recorded request body is written to a __files/ file and the mapping's request matcher references it by name. The matcher form is chosen from the request's content type, so the externalised body still matches on replay exactly as an inline one would:

• a JSON request body produces {"equalToJson":{"bodyFileName":"…"}};
• an XML request body produces {"equalToXml":{"bodyFileName":"…"}};
• a binary request body produces {"binaryEqualTo":{"bodyFileName":"…"}} (the file is written byte-for-byte);
• any other body (plain text, or a request with no Content-Type) produces {"equalTo":{"bodyFileName":"…"}}.

Two kinds of request body are always kept inline, regardless of the checkbox, and are never externalised:

• an empty request body; and
• a multipart request body.

When the option is off (the default), request bodies are recorded inline exactly as in earlier releases, so existing recording workflows are unchanged.

The checkbox controls a single recording session. To change the default state of the checkbox for everyone (for example to have it ticked out of the box), set the trafficparrot.http.recording.request.externaliseToFile property in trafficparrot.properties. The property only seeds the checkbox's initial state: whatever the checkbox is set to when you start recording is what takes effect for that session.

Traffic Parrot can automatically rewrite URLs in response bodies during recording. This is useful when recording from production systems where responses contain absolute URLs that need to be changed to point to your test environment.

Configuration

Configure URL rewriting by setting the trafficparrot.http.recording.response.rewrite.regex property in trafficparrot.properties. The format is:

trafficparrot.http.recording.response.rewrite.regex=pattern->replacement,pattern2->replacement2

Examples

Simple string replacement:

trafficparrot.http.recording.response.rewrite.regex=https://api.example.com->http://localhost:8081

Regex with capture groups:

trafficparrot.http.recording.response.rewrite.regex=(https?://[^/]+)(.*)->http://localhost:8081$2

Multiple rules:

trafficparrot.http.recording.response.rewrite.regex=(https://api.example.com)(.*)->http://localhost:8081$2,(https://cdn.example.com)(.*)->http://localhost:8082$2

How it works

• URL rewriting is applied only during recording mode
• The live response the client receives is transformed by trafficparrot.http.recording.response.rewrite.regex; the body written to the saved mapping file is transformed separately by trafficparrot.http.recording.storage.rewrite.regex (often in the opposite direction, to preserve Handlebars templates)
• Works with gzip, brotli, and uncompressed responses
• Only text-based response bodies are processed (JSON, XML, HTML, etc.)

The properties above rewrite response bodies. To rewrite the request body as it is recorded, set trafficparrot.http.recording.request.storage.rewrite.regex. The usual reason is to keep a credential, token or item of personal data out of your recorded traffic files.

It is empty by default, so nothing is rewritten until you configure it. The format is the same as the other rewrite properties:

trafficparrot.http.recording.request.storage.rewrite.regex=("apiKey"\\s*:\\s*")[^"]*->$1***REDACTED***

Record a POST whose body is {"apiKey":"live-key-1234","orderId":42} and the saved mapping matches on {"apiKey":"***REDACTED***","orderId":42} instead. The real key is in no traffic file Traffic Parrot writes. It is still in the per-request entry of the log while request logging is on, which is the default: set trafficparrot.virtualservice.httpRequestLoggingDisabled=true while you record if that matters, or treat the log file as sensitive.

What is covered

• Rules run on every non-empty request body that is not multipart, whatever its content type. Bodies are decoded using the charset from the request's Content-Type header, defaulting to UTF-8.
• multipart/form-data request bodies are recorded unchanged, with no warning, even when rules are set. Traffic Parrot stores a multipart body inline without passing it through the rewrite. Do not rely on this property to scrub a file upload.
• Empty request bodies are left alone, as there is nothing to match.
• A body that no rule matches is written byte-for-byte as received, because nothing is decoded. A body that a rule does match is decoded and re-encoded in full, so every byte that is not valid in the request's character set becomes the replacement character U+FFFD, anywhere in the body and not only where the rule matched. If an endpoint you record can also carry a file upload, keep the upload out of the recording rather than relying on the pattern to miss it: give the Recording URL a path so that only requests under that path are saved. The content type filter matches the response's Content-Type, not the request's, so it does not keep out an upload whose response is JSON.
• This property is independent of trafficparrot.http.recording.storage.rewrite.regex, which still applies to response bodies only. Set both if you want both sides scrubbed.
• Rules apply whether the request body is stored inline in the mapping or externalised to a __files/ file. See Externalising request bodies to files.

Writing the rules

• Double every backslash. trafficparrot.properties is a Java properties file, which reads a single \s or \d as a bare s or d and so changes what the rule matches without reporting anything. Write \\s and \\d, as the example above does.
• A comma separates rules and an arrow separates a pattern from its replacement, so neither can appear inside a pattern. A range quantifier such as \\d{12,19} splits into two rules; write \\d{12}\\d* (twelve or more digits) instead, or spell the optional digits out as repeated \\d? if the upper bound matters.
• Keep the surrounding quotes when redacting a JSON string value. Capture the field name and the opening quote, as the example above does, so the result is still valid JSON. If a rewrite leaves a JSON request body invalid, the exchange is not recorded at all, the client receives a 500 from the proxy even though the backend has already handled the request, and an error naming the URL and the rules is written to the log.
• A malformed rule stops Traffic Parrot from starting, and the error names the rule at fault. This is deliberate: the property exists to keep values out of recordings, so a rule that cannot be parsed must not be skipped silently while recording continues. The same applies to trafficparrot.http.recording.response.rewrite.regex and trafficparrot.http.recording.storage.rewrite.regex.

There are two unrelated things people call "redirects" in Traffic Parrot, and they are controlled separately. Read this first so you know which one applies to you.

1. Serving a stubbed 3xx (always available)

You can define a mapping that returns a 301 or 302 response with a Location header. This is standard mock behaviour: Traffic Parrot returns the redirect to the client exactly as you defined it. This is always available and is not affected by the trafficparrot.virtualservice.enableRedirectHandling property.

For example, a mapping that redirects clients calling /old-path to /new-path:

{
  "request": {
    "method": "GET",
    "url": "/old-path"
  },
  "response": {
    "status": 302,
    "headers": {
      "Location": "/new-path"
    }
  }
}

Clients calling /old-path receive HTTP/1.1 302 Found with Location: /new-path, just like any other stubbed response.

2. Following a backend redirect in proxy/record mode (property-controlled)

When Traffic Parrot is recording by proxying to a real backend and that backend returns a 3xx redirect (e.g. 302 with a Location header), the trafficparrot.virtualservice.enableRedirectHandling property decides whether the redirect itself is recorded, or whether Traffic Parrot follows it and records the final response instead.

The property defaults to false. It is read once at startup (it is a server setting, not a per-recording-session option, so there is no toggle on the Record page). Set it in trafficparrot.properties:

trafficparrot.virtualservice.enableRedirectHandling=true

The following table summarises the recording behaviour for a backend that answers a request with 302 + Location, where the target of the Location returns a terminal 200:

When to use each

• Leave enableRedirectHandling at its default false when you want the redirect itself reproduced in your virtual service, so replay returns the 302 + Location exactly as the backend did.
• Set it to true for transparent recording where you only care about the final response: Traffic Parrot follows the redirect hop while recording and saves only the terminal response, so your mocks contain no redirect.

Mapping cleanup is a page in the Traffic Parrot admin UI. To open it, go to the HTTP menu in the top navigation bar and choose Cleanup. It rewrites a scenario's HTTP mappings into a clean, convention-following form.

Recording a real API (or importing a HAR) typically produces dozens of near-identical per-request stubs (one each for /users/1, /users/2, /users/3 and so on), each carrying noisy response headers and over-strict matchers. The result is hard to maintain. Cleanup consolidates these stubs into a small set of pattern-based mappings.

Cleanup is non-destructive. It never modifies the source scenario. The cleaned mappings are written to a new scenario named by the targetScenario parameter; the original scenario is left untouched. The new scenario carries the HTTP mappings and __files only, so it is not a complete copy of the source: see What the cleaned scenario does not carry.

Cleanup is convention-only today. The standard cleanup applies a fixed set of conventions: there are no configurable parameters or rule toggles. The only input is the target scenario name.

For a step-by-step walkthrough, see the tutorial How to clean up recorded HTTP mappings.

Standard cleanup applies these conventions to every mapping in the scenario:

The last two transforms (making response URLs portable and adding response templates) correspond to the urlsReplaced and templatesAdded result counters described below.

Cleanup copies the source scenario's HTTP mappings and its __files directory into the new scenario, and nothing else. A scenario can also hold mappings for other protocols, in directories beside mappings: jms-mappings, ibm-mq-mappings, file-mappings, grpc-mappings and websocket-mappings. None of those is copied, and neither is anything else the scenario holds, such as its data directory.

That is deliberate. The conventions in the table above are HTTP conventions: URL patterns, response headers, response templating. They have no meaning for a JMS or a gRPC mapping, so cleanup leaves those mappings where they are instead of rewriting them.

Nothing is lost. Cleanup never modifies the source scenario, so its messaging, gRPC and WebSocket mappings are still there, exactly as they were. What this guards against is a lossy migration: treating the cleaned scenario as a complete replacement for the recording it came from. Promote only the cleaned scenario to the next environment and the other protocols' mappings do not travel with it; delete the original once you have switched to the cleaned one and they are gone. Keep the source scenario, or copy the directories you still need into the cleaned one yourself.

Cleanup reports the mapping directories it did not copy, so you do not have to go looking for them. Each entry names the directory and how many files stayed behind, for example jms-mappings (3 file(s)) - not copied into the cleaned scenario. The count is of files, not mappings: one mapping file can hold several mappings.

Only a directory that actually holds files is reported. Traffic Parrot creates all of these directories in every scenario, so an empty one says nothing about whether the scenario uses that protocol, and cleaning up an HTTP-only scenario reports nothing at all.

The report is informational: it is a warning, not an error. The cleanup succeeded, and the errors field stays empty. You see it in the warning panel on the cleanup page, in the droppedMappingDirectories field of the REST API response, and in the output of the cleanup_mappings MCP tool.

1. From the HTTP menu in the top navigation bar, choose Cleanup. The page shows the current scenario and how many mappings it contains.
2. The Target scenario name field is pre-filled with a suggested name (the current scenario name with _cleaned appended). Change it if you like; it has to be a name inside the scenarios directory, so a name that climbs out of it with .. or an absolute path is refused, and the page says why rather than writing anything.

3. Click Apply Standard Cleanup.
4. The Results panel shows the before/after counts; click switch to the cleaned scenario to start using the simplified mappings.

The results panel reports six counters: Original Mappings, Cleaned Mappings, Mappings Consolidated, Headers Stripped, Body Patterns Removed, and Files Renamed.

When cleanup consolidates near-duplicate mappings (same method and URL pattern, only the smallest-body mapping is kept), a warning panel lists exactly which mappings were consolidated away, for example GET /api/store/[0-9]+/items(.*) (response: body 61 chars). Cleanup is no longer silent about what it drops: review this list before discarding the original recording, in case a consolidated-away mapping carried a response you still need.

Above the counters, a warning panel headed Mappings the cleaned scenario does not serve appears when the source scenario also holds mappings for another protocol. It names each mapping directory that was not copied and how many files stayed behind, for example jms-mappings (3 file(s)) - not copied into the cleaned scenario. The panel is hidden when there is nothing to report, which is the normal case for an HTTP-only scenario. See What the cleaned scenario does not carry.

Below the summary counters, expand Per-endpoint detail to see, for every recorded endpoint, exactly what cleanup did to it. Each row is tagged Kept, Consolidated, Headers stripped, Body match removed, URL pattern applied, Response template added, or Removed. A consolidation groups block underneath makes the grouping explicit, listing which specific recorded URLs were collapsed into each surviving pattern. Because cleanup is intentionally lossy, this view lets you confirm what changed before you discard the original recording.

The cleanup endpoints run on the Traffic Parrot GUI/admin port. You can use them to inspect the current scenario and run cleanup programmatically.

Send a GET request to /api/http/cleanup/mappings to return the current scenario and its mapping count:

curl http://localhost:8080/api/http/cleanup/mappings

The response includes a short preview list (sampleMappings) of up to 10 file names:

{
  "scenario": "Default",
  "mappingCount": 42,
  "sampleMappings": ["get-users1-….json", "get-users2-….json"]
}

Send a GET request to /api/http/cleanup/suggestedName to return a suggested target name. The response is a JSON object, not a bare string:

curl http://localhost:8080/api/http/cleanup/suggestedName

{
  "suggestedName": "Default_cleaned",
  "currentScenario": "Default"
}

Send a POST request to /api/http/cleanup/execute?targetScenario=<name> to run the standard cleanup and write the result to the new scenario. The targetScenario query parameter is required; there is no request body.

curl -X POST "http://localhost:8080/api/http/cleanup/execute?targetScenario=cleaned-api"

The value names a new scenario inside the instance's scenarios directory. A name that would resolve outside that directory (one climbing out of it with .., or an absolute path) is rejected, as are a blank name, the name of the current scenario, and the name of a scenario that already exists. A rejected name returns HTTP 400 with the reason in an error field, and nothing is written:

{
  "error": "Target scenario '../cleaned-api' must be a name inside the scenarios directory. Please provide a scenario name without '..' or an absolute path."
}

On success, the response is the cleanup result:

{
  "targetScenario": "cleaned-api",
  "originalMappingsCount": 42,
  "cleanedMappingsCount": 18,
  "mappingsConsolidated": 20,
  "headersRemoved": 15,
  "bodyPatternsRemoved": 8,
  "filesRenamed": 5,
  "urlsReplaced": 12,
  "templatesAdded": 0,
  "consolidatedAwayMappings": [
    "GET /api/store/[0-9]+/items(.*) (response: body 61 chars)",
    "GET /api/store/[0-9]+/items(.*) (response: body 36 chars)"
  ],
  "removedEndpoints": [],
  "droppedMappingDirectories": [
    "jms-mappings (3 file(s)) - not copied into the cleaned scenario",
    "grpc-mappings (1 file(s)) - not copied into the cleaned scenario"
  ],
  "endpointDispositions": [
    { "endpoint": "GET /users/[0-9]+", "actions": ["CONSOLIDATED", "HEADERS_STRIPPED"] },
    { "endpoint": "GET /health", "actions": ["KEPT"] }
  ],
  "consolidationGroups": [
    {
      "pattern": "GET /users/[0-9]+",
      "consolidatedUrls": [
        "GET /users/123 (response: body 61 chars)",
        "GET /users/456 (response: body 36 chars)"
      ]
    }
  ],
  "errors": []
}

The result fields are:

• originalMappingsCount / cleanedMappingsCount: the mapping count before and after cleanup.
• mappingsConsolidated: stubs merged into patterns or deduplicated.
• headersRemoved: mappings whose response headers were trimmed.
• bodyPatternsRemoved: request body matchers that were dropped.
• filesRenamed: response body files that were renamed.
• urlsReplaced: response bodies where an absolute URL was replaced with {{request.baseUrl}}.
• templatesAdded: responses where a dynamic {{request.path.[…]}} template was inserted.
• consolidatedAwayMappings: the identity of each near-duplicate mapping that consolidation removed, formatted as METHOD url-pattern (response: body N chars | file <name> | none). Empty when nothing was consolidated.
• removedEndpoints: any endpoint (method and normalised path) that was mapped before cleanup but has no mapping afterwards. Normally empty, because consolidation always keeps one survivor per endpoint; it is a safety guard that only fires if a transform were to remove an endpoint entirely.
• droppedMappingDirectories: the mapping directories the source scenario holds that the cleaned scenario does not carry, one entry per directory, formatted as <directory> (N file(s)) - not copied into the cleaned scenario. A directory that holds no files is not listed, so this is empty for an HTTP-only scenario. It is a warning rather than an error: the cleanup itself succeeded. See What the cleaned scenario does not carry.
• endpointDispositions: one entry per recorded endpoint, each { "endpoint": "GET /users/[0-9]+", "actions": […] }, where every action is one of KEPT, URL_PATTERN_APPLIED, HEADERS_STRIPPED, BODY_PATTERN_REMOVED, TEMPLATE_ADDED, CONSOLIDATED, or REMOVED. This is the per-endpoint view behind the Per-endpoint detail table in the GUI.
• consolidationGroups: one entry per surviving pattern, each { "pattern": "GET /users/[0-9]+", "consolidatedUrls": […] }, listing which recorded URLs were collapsed into that pattern. Empty when nothing was consolidated.
• errors: any mappings that could not be cleaned (empty on success).

The standard cleanup is convention-driven and is not currently configurable.

Endpoint coverage tells you, for the current scenario, which HTTP endpoints Traffic Parrot knows about and how the real traffic it has seen lines up with your stub mappings. Every endpoint is classified as exactly one of three states:

• Mapped: at least one stub mapping exists for the endpoint and it has been observed in traffic.
• Observed but unmapped: requests for the endpoint appear in the request journal but no stub matched them, so they fell through to the unmatched HTTP 900 No responses matched response. These are gaps in the virtual service.
• Mapped, never observed: a stub exists but no request has exercised it yet. This is a candidate for cleanup (possibly an obsolete mapping).

Endpoint coverage complements Mapping cleanup: after consolidating mappings, coverage answers the question “did consolidation accidentally drop an endpoint that real traffic still hits?”. An endpoint that becomes Observed but unmapped after a cleanup is a regression signal. Coverage and cleanup use the same endpoint normalisation (URL patterns collapse concrete URLs such as /orders/123 and /orders/456 into a single endpoint), so the two pages always agree on what counts as one “endpoint”.

“Observed” is derived from the HTTP request journal: the serve events the virtual service records as requests arrive. Traffic Parrot's HTTP engine is based on WireMock, so this is the same request journal described in the WireMock verifying / request-journal documentation.

Open the coverage page from the top navigation bar: go to the HTTP menu and choose Coverage (direct URL /http/coverage.html). It is available when virtual services are enabled.

The page header shows a summary line (for example 3 endpoints: 2 mapped, 1 observed but unmapped, 1 mapped but never observed) and a caveat line stating how much traffic the figures are based on, for example Based on 2 requests in the request journal. A Refresh button reloads the data, and the table refreshes live.

The table has six columns, in this order:

The table is searchable using the Search: box, and each column is sortable.

How to read it:

• Red Observed but unmapped rows are gaps: traffic with no stub behind it.
• Grey Mapped, never observed rows are cleanup candidates: stubs no request has exercised.
• A non-zero Unmatched count on an otherwise-mapped endpoint means some requests to that path did not match any of its stubs, for example matchers that are stricter than the real traffic. See the request log for the per-request near-misses.

Send a GET request to /api/http/coverage on the Traffic Parrot GUI/admin port to return the same data as JSON for the current scenario. It takes no parameters and is always available, independent of whether a recording is in progress.

curl http://localhost:8080/api/http/coverage

A representative response (trimmed to two endpoints):

{
  "endpoints": [
    { "method": "GET", "path": "/api/orders(.*)",        "mapped": true,  "mappedCount": 1, "observedCount": 1, "unmatchedCount": 0 },
    { "method": "GET", "path": "/api/unmapped-thing(.*)", "mapped": false, "mappedCount": 0, "observedCount": 1, "unmatchedCount": 1 }
  ],
  "totalEndpoints": 3,
  "mappedEndpoints": 2,
  "observedButUnmappedEndpoints": 1,
  "mappedButNeverObservedEndpoints": 1,
  "totalRequestsInJournal": 2
}

The top-level aggregates are totalEndpoints, mappedEndpoints, observedButUnmappedEndpoints, mappedButNeverObservedEndpoints, and totalRequestsInJournal. Each entry in endpoints carries method, path, mapped, mappedCount, observedCount, and unmatchedCount.

/api/http/coverage is the standalone, always-available endpoint-coverage view: you can call it at any time, for any scenario. It shares this same per-endpoint model with the Live coverage REST API (/api/http/currentRecordingCoverage), which wraps the same data in a top-level recording flag but is scoped to an in-progress HTTP recording session; this endpoint is the live monitor during a recording, whereas /api/http/coverage is the always-on view.

Instead of pasting a response body inline into a mapping, you can keep the body in a separate file and point the mapping at it with the bodyFileName attribute. This is useful when you want to:

• Serve large or binary bodies (a PDF, an image, a multi-kilobyte JSON or XML document) that would clutter the mapping if inlined
• Reuse the same body across more than one mapping
• Keep your mappings/*.json files small and readable, with the payload stored alongside them
• Edit the response body in your editor of choice, or generate it from a build step, without touching the mapping

The same bodyFileName / __files convention also externalises a request body matcher; see Request body file.

Coming from WireMock? Traffic Parrot is a WireMock-compatible virtual service, so the bodyFileName response attribute and the __files directory convention work unchanged. Mappings you already have that use bodyFileName (including those you import as a WireMock ZIP) load and serve in Traffic Parrot without modification. This page is the Traffic Parrot equivalent of WireMock's "response from a file" / "body file location" documentation.

A bodyFileName is resolved relative to the __files directory that sits next to the directory the mapping lives in (a sibling of mappings), not relative to the mapping file itself. A mapping in mappings/foo.json with "bodyFileName": "data.json" loads __files/data.json.

The default layout looks like this:

trafficparrot-x.y.z/
├── mappings/
│   └── foo.json          ← "response": { "bodyFileName": "data.json" }
└── __files/
    └── data.json         ← the response body that is returned

Nested paths are allowed. A bodyFileName may contain forward-slash subdirectories, which resolve under __files:

"response": { "bodyFileName": "sub/data.json" }   →   __files/sub/data.json

Text bodies are read as UTF-8. Binary bodies (images, PDFs, archives) are read as bytes, so the file is returned exactly as stored; see Binary file in response and Serving static images for binary examples.

This walkthrough creates a stub for GET /greeting whose response body comes from a file.

1. Create the body file __files/body.txt with the text you want to return:

   Hello from a file-backed response!

2. Create the mapping mappings/greeting.json that references it through bodyFileName:

   {
     "request": {
       "method": "GET",
       "url": "/greeting"
     },
     "response": {
       "status": 200,
       "bodyFileName": "body.txt",
       "headers": {
         "Content-Type": "text/plain"
       }
     }
   }

3. Start Traffic Parrot (or, if it is already running, the mapping is picked up automatically). Then call the stub on the virtual service port (trafficparrot.virtualservice.http.port, 8081 by default):

   curl http://localhost:8081/greeting

   The response body is the contents of __files/body.txt:

   Hello from a file-backed response!

When you open this mapping in the Add/Edit page, a note above the response body field tells you the body is served from a file rather than typed inline, and the field is pre-filled with the file's contents so you can edit it in place (see Editing a file-backed response in the UI).

When you open a mapping whose response uses bodyFileName on the Add/Edit page, Traffic Parrot shows an information note directly above the response body field reading Response from file: followed by the file name. The response body field is pre-filled with the file's current contents (loaded from __files) and is editable in place, so you can change a file-backed response without leaving the form.

When you edit the body and click Save, Traffic Parrot writes your changes back to the body file in __files (through its file source, so the in-memory mapping cache stays in sync). The mapping keeps its bodyFileName reference; only the file's contents change. If the file cannot be read or written, see Troubleshooting: missing body file.

When the body file is an image (PNG, JPEG, GIF, SVG, WebP, BMP, ICO, APNG or AVIF), Traffic Parrot renders the image inline above the Response from file note, so you can see at a glance which image the mapping serves without opening the Static Files browser or the file on disk. Because an image is binary, the body field stays read-only and is edited on disk; the preview shows what the response is, while the file name in the note tells you which file to change. The same inline rendering is used in the Static Files browser preview. Non-image binary files (a PDF, an archive) keep the read-only body field with no preview.

If the same body file is referenced by more than one mapping, Traffic Parrot shows a warning above the field. Editing a shared file changes the response for every mapping that uses it, so the warning names how many other mappings are affected.

You can switch a response between inline and file-backed from the same form:

• Switch to inline detaches the bodyFileName and keeps the current body as an inline response. The file on disk is left untouched.
• Switch to file externalises an inline body to a file in a single click: the response body becomes file-backed straight away, a file name field appears pre-filled with a suggested name (for example responses/<name>.json), and the Response from file note is shown. There is no separate confirm step: clicking Save writes the body to that file under __files using the current file name (the suggested default, or whatever you typed) and points the mapping at it. The file name field stays editable; clearing it keeps the body inline.

When the file name field is shown (after Switch to file), a Choose an existing file... dropdown appears beside it, listing the files already in the scenario's __files directory (including any nested responses/<name>.<ext> files). Selecting an entry fills the file name field with that path, so you can point the response at an existing body file without typing, useful when you want to reuse a payload that another mapping already serves. The dropdown is a convenience only: the file name field stays editable, so typing the name by hand still works exactly as before, and clearing the field keeps the body inline. Picking an existing file does not load its contents into the editor; just as when you type the file name, clicking Save writes the response body currently in the editor to that file, so if the body differs from the file's existing contents Save overwrites the file.

A Manage files link in the response body area opens the Static Files browser in a new browser tab, so you can list, create, rename or delete the files in the __files directory without leaving, or losing, the mapping you are editing. The link is always shown, whether the response is inline or already file-backed.

In the HTTP Add/Edit mappings list, the Response body column normally shows the response body content itself, whether the body is inline, recorded, or kept in a text body file whose contents are read as text. For these rows the column reads exactly the same as if the body had been pasted inline.

A binary body file is different: its bytes are not text (an image, a PDF, an archive; see Templating vs byte-for-byte), so there is no readable body to show in the column. For such a row the Response body column shows the bodyFileName path the body is served from (for example responses/sample-image.jpg) instead. Long paths are truncated to fit the column; hover over the cell to see the full path in a tooltip.

When the binary body file is an image (PNG, JPEG, GIF, WebP, BMP, ICO, APNG or AVIF), hovering the cell goes one step further: instead of the full-path tooltip it pops a small rendered preview of the image itself, so you can recognise which picture a mapping serves without opening the editor or the file on disk. The cell still shows the truncated bodyFileName path; the preview only replaces what the hover shows. The image is loaded on the first hover and capped to a small thumbnail; if it cannot be read the cell falls back to the plain full-path tooltip. This is the same image that is rendered inline when you open the mapping (see Editing a file-backed response in the UI). Non-image binary rows (a PDF, an archive) keep the full-path tooltip.

This lets you tell at a glance which mapping serves which binary file without opening each one. Only binary file-backed rows change; inline, recorded, and text file-backed rows show their body content as before.

If you organise mappings into scenarios, each scenario has its own __files directory alongside its mappings directory. A bodyFileName on a scenario's mapping resolves against that scenario's __files:

scenarios/
└── my-scenario/
    ├── mappings/
    │   └── greeting.json   ← "bodyFileName": "body.txt"
    └── __files/
        └── body.txt        ← scenario-scoped response body

This keeps each scenario self-contained, so the bodies a scenario depends on travel with the scenario when you copy or share it.

By default the files root is the Traffic Parrot working directory (file:.), so body files live under __files in the Traffic Parrot install directory (and under each scenario for scenario-scoped mappings). If you want your body files and mappings to live somewhere else (for example a shared data directory under version control, or a mounted volume), set the files root with a property in trafficparrot.properties:

trafficparrot.virtualservice.trafficFilesRootUrl=file:/path/to/data

The value is a file: URL pointing at the directory that contains your mappings and __files directories. Its default is file:. (the current working directory). With the files root pointed at /path/to/data, a mapping's bodyFileName resolves under /path/to/data/__files (or the matching scenario's __files). See the property reference for trafficparrot.virtualservice.trafficFilesRootUrl for the full description and further examples.

A body file is not limited to a fixed payload. It can contain the same Handlebars templating helpers you can use in an inline response body, so a single file-backed response can vary with the incoming request. This combines the readability of an external file with the flexibility of dynamic responses.

For example, a mapping that echoes back a list of ids from the request body:

{
  "request": {
    "urlPath": "/requestJsonPathList",
    "method": "ANY",
    "bodyPatterns": [ {
      "matchesJsonPath": "$.items"
    } ]
  },
  "response": {
    "status": 200,
    "bodyFileName": "requestJsonPathList.json"
  }
}

with the body file __files/requestJsonPathList.json using Handlebars to iterate over the request's items array:

{
    "type": "response",
    "items": [
        {{#each (jsonPathList request.body '$.items') }}
        { "id": {{ jsonPath this '$.id' }}, "status": "OK" }{{#unless @last}},{{/unless}}
        {{/each}}
    ]
}

A ready-to-run version of this example ships in the examples project. For the full list of helpers and request data you can reference from a templated body (whether inline or file-backed), see Generate responses and Use request data in response.

When HTTP Handlebars response templating is enabled (the default, trafficparrot.http.handlebars.enabled=true), Traffic Parrot classifies each bodyFileName response as binary or text and treats the two differently:

• Binary body files are served byte-for-byte and are never run through Handlebars. A body file is treated as binary when there is positive binary evidence: either the response Content-Type is a binary type (image/*, audio/*, video/*, font/*, application/pdf or application/octet-stream), or the body file has a known binary extension (jpg, png, gif, pdf, zip, woff, mp4, …). The bytes are returned exactly as stored, so an image or PDF is never corrupted by being read and re-encoded as text.
• Everything else is treated as text and rendered through Handlebars. Text formats such as JSON, XML, plain text and SVG (SVG is XML, so it is text) are templated as before, and so is any body file with no recognised binary Content-Type and no binary extension.

This means a template body file with no Content-Type (for example a .hbs, .tpl or .md file that contains {{…}} placeholders) is still rendered, while a binary file such as sample-image.jpg is served byte-for-byte whether or not it has an image/jpeg Content-Type. No per-mapping configuration is needed for either case.

Forcing byte-for-byte for a file Traffic Parrot cannot auto-detect

If you serve a binary file with an unusual extension and no Content-Type, Traffic Parrot has no positive binary evidence, so it defaults to treating the file as text and running it through Handlebars (which can corrupt it). To force such a file to be served byte-for-byte, do either of the following:

• Set "disableBodyFileTemplating": true in the response's transformerParameters, which turns off templating for that response:

  {
    "response": {
      "status": 200,
      "bodyFileName": "payload.dat",
      "transformerParameters": {
        "disableBodyFileTemplating": true
      }
    }
  }

• Give the response a binary Content-Type (for example application/octet-stream) in its headers, which Traffic Parrot recognises as positive binary evidence:

  {
    "response": {
      "status": 200,
      "bodyFileName": "payload.dat",
      "headers": {
        "Content-Type": "application/octet-stream"
      }
    }
  }

Handlebars partials referenced with {{>name}} (for example a .hbs partial) are always rendered as part of templating; they are not bodyFileName responses, so this binary/text classification does not apply to them.

Per-file-type reference

The table below shows how some common body files are classified.

If a mapping references a bodyFileName that does not exist under the resolved __files directory, Traffic Parrot reports the problem rather than silently returning an empty body: the matched request returns an HTTP 500 error and a FileNotFoundException naming the expected path is logged. Check that:

• The file name in bodyFileName matches the file on disk exactly, including case and any subdirectory prefix
• The file is in the __files directory that is a sibling of the mapping's mappings directory (or the scenario's __files for scenario-scoped mappings)
• If you have set trafficparrot.virtualservice.trafficFilesRootUrl to a non-default location, the __files directory is under that location

The same "resolve relative to the sibling __files" rule also applies when you externalise a request body to a file for matching, so request- and response-side body files live side by side under the same __files directory. (A missing request body file is reported earlier still, when the mapping is loaded, rather than at request time.)

• Binary file in response: return a PDF or other binary file from __files
• Serving static images: serve images from __files, including with URL patterns
• Dynamic responses: the Handlebars helpers you can use inside a templated body file
• Import and export HTTP mappings: import WireMock ZIP mappings that use bodyFileName unchanged
• Editing a file-backed response in the UI: the "Response from file" note and editable body on the Add/Edit page

The Static Files browser is an admin-UI page for the __files directory that holds your file-backed response bodies. Instead of dropping to the filesystem, you can list, preview, create, edit, rename, upload and delete the files that mappings refer to with bodyFileName, all from the browser, against the same files Traffic Parrot serves at run time.

Open it from the Traffic Parrot admin UI: HTTP menu → Static Files (the page is served at /http/files.html). This is a different menu item from the top-level Files entry, which is the file-based message protocol.

The browser lists the files under the active scenario's __files directory, the same directory a mapping's bodyFileName resolves against at serve time. The active scenario name is shown at the top of the page, so the file list always matches the files the current scenario's mappings actually read from. If you switch scenarios, the browser shows that scenario's own __files.

Files in sub-directories of __files are grouped under expandable folder rows rather than listed by their full relative path. Each folder has a caret you can click to collapse or expand its contents (all folders are expanded by default), and each file is labelled by its own name and indented under its folder. When the directory has no files, the browser shows an empty-state message instead of the table.

The Reference column tells you, at a glance, whether each file is still in use. A file is marked referenced when at least one mapping in the active scenario serves it as a response body (the mapping's response.bodyFileName equals the file's relative name), and orphaned when no mapping serves it. This helps you spot body files that are safe to clean up and ones that a response depends on.

The badge inspects only response references. A file used as a request body matcher can therefore show as orphaned even though a mapping still depends on it. The delete and rename safety checks are more complete than the badge (they also consider request-body references), so a file the badge calls orphaned can still be refused for deletion or rename when it is genuinely in use.

Click a file to preview its content in the right-hand Preview pane:

• Text, JSON and XML files are shown as text; JSON and XML are pretty-printed so they are easy to read.
• Image files (PNG, JPEG, GIF, SVG, WebP, BMP, ICO, APNG and AVIF) are rendered inline, so you can see at a glance which image a mapping serves.
• Any other binary file shows a short note with its content type and size (for example Binary file - application/octet-stream, 2.3 KB) rather than dumping raw bytes.

You can drag the divider between the Files and Preview panes to widen either side to suit the file you are reading (the split stays between 20% and 80% so neither pane collapses), and double-click the divider to snap back to the default split. The split you choose is remembered in your browser across page reloads. On narrow screens the panes stack vertically.

Click New file to create a text file under the active scenario's __files directory without leaving the page. Enter a relative file name (for example responses/example.json; any forward-slash sub-directories are created for you) and the file content, then click Create. The new file appears in the list with an orphaned badge until a mapping references it. A name that already exists, or one that would escape the __files directory (an absolute path or a path containing ..), is rejected with an error rather than silently overwriting or escaping the root.

Click Upload to add an existing file from your computer to the __files directory. Both text and binary files are accepted and their bytes are stored verbatim, so an uploaded image is listed and rendered inline in the preview pane, and any other uploaded binary file is listed with its content type and size. Uploading a file whose name already exists is refused so you never silently overwrite another file, and unsafe file names are rejected just as for New file.

The upload has a configurable size limit, set by the trafficparrot.multipart.upload.max.size.mb property (default 10 MB, shared by all admin-UI file uploads). A file larger than the limit is rejected with HTTP 413 and the message File exceeds the maximum upload size of N MB shown in the upload form. The limit applies only to file uploads through the admin UI and does not affect recorded traffic or mapping sizes.

Select a text file and click Edit to replace the preview with an editable text area and Save / Cancel buttons. Saving overwrites the file's content in the __files directory. If the file is referenced by one or more mappings, a warning is shown above the editor, because saving changes the response for every mapping that serves it. Image and other binary files stay read-only and are edited on disk.

This is the same body file you can edit from the Add/Edit mapping form; see Editing a file-backed response in the UI. The Static Files browser lets you edit the file directly, even when you do not have the mapping open.

Each row has a Rename button that reveals an inline editable target-name field with Confirm / Cancel (no browser pop-up). Renaming a file moves it to the new name, creating any intermediate sub-directories the new name implies. Renaming onto a name that already exists is refused, so you never silently overwrite another file.

Rename is reference-safe: if the file still backs a mapping, the rename is refused, a banner lists the referencing mapping(s), and the file keeps its old name, because renaming it would leave the mapping pointing at a name that no longer exists. The reference check covers both a file used as an HTTP response body (response.bodyFileName) and one used as a request body matcher. To rename a referenced file, first remove or repoint the mapping(s) that use it; Traffic Parrot does not automatically rewrite the referencing mappings to the new name.

Each row has a Delete button that reveals an inline Confirm / Cancel step (no browser pop-up). Confirming the deletion of an unreferenced file removes it from disk and the row disappears.

Delete is reference-safe: if the file still backs a mapping, the deletion is refused, a banner lists the referencing mapping(s), and the file stays listed, so you cannot silently break a response or a request matcher by deleting the file it serves. As with rename, the check considers both response.bodyFileName and request-body references, so a file the orphaned badge reports as unreferenced can still be blocked from deletion when it is genuinely in use. To delete a referenced file, first remove or repoint the mapping(s) that use it.

When a browser application (a single-page app, a front-end calling your mock from fetch/XMLHttpRequest) calls a mocked endpoint on a different origin, the browser enforces the same-origin policy: it first sends a CORS preflight (an HTTP OPTIONS request) and then expects the response to carry Access-Control-* headers, notably Access-Control-Allow-Origin. If those headers are missing, the browser blocks the call and you see a “blocked by CORS policy” error in the console, even though the request reached the mock and a response came back.

Traffic Parrot's virtual service is CORS-friendly out of the box. There is no flag to set and no manual stub to add: for every path on the virtual service (the port your browser app calls), Traffic Parrot automatically answers the CORS preflight and reflects the request's Origin into Access-Control-Allow-Origin on the actual response. Your browser clients are not blocked by the same-origin policy with no configuration at all.

This section is about stub / mock-response CORS: the CORS headers the virtual service sends to the browser apps that call your mocks. It is independent of the Traffic Parrot management API and web UI.

The behaviour below applies to the whole virtual service. It is global, not per-stub: an un-stubbed path returns the same CORS preflight as a stubbed one, so you do not have to remember to add CORS to each mapping.

• CORS preflight is answered automatically. An OPTIONS request to any path that carries an Origin and an Access-Control-Request-Method header (what a browser sends as a preflight) is answered with HTTP 200 and the following headers:
  • Access-Control-Allow-Origin: set to the request's Origin, reflected back verbatim (for example a request Origin: http://localhost:3000 yields Access-Control-Allow-Origin: http://localhost:3000)
  • Vary: Origin
  • Access-Control-Max-Age: 60
  • Access-Control-Allow-Methods: the methods OPTIONS, GET, POST, PUT, PATCH and DELETE
  • Access-Control-Allow-Headers: the request headers X-Requested-With, Content-Type, Accept, Origin and Authorization
  The Access-Control-Allow-Methods and Access-Control-Allow-Headers values are returned as a set: all of the entries above are present, but do not rely on a fixed comma order, which is not guaranteed between runs.
• The actual response reflects the Origin too. A real request (for example a GET) that carries an Origin header gets Access-Control-Allow-Origin: <the reflected Origin> and Vary: Origin added to its response, even when the matched stub sets no CORS headers itself. You do not have to add CORS headers to your mappings for browser clients to work.
• No Origin, no reflection. A request that does not carry an Origin header (a server-to-server call, a curl without -H 'Origin: …') gets only Vary: Origin and no Access-Control-Allow-Origin. This is standard CORS: the header is reflected only when the caller is a browser sending an Origin.

You can reproduce the browser's two-step CORS exchange with curl against any path on the virtual service port (trafficparrot.virtualservice.http.port, 8081 by default). The path does not need a stub for the preflight to be answered; replace /your/path with one of your mapped endpoints.

1. The preflight. This is the OPTIONS request a browser sends first:

curl -i -X OPTIONS http://localhost:8081/your/path \
  -H 'Origin: http://localhost:3000' \
  -H 'Access-Control-Request-Method: GET'

Traffic Parrot answers it automatically (abbreviated):

HTTP/1.1 200 OK
Access-Control-Allow-Origin: http://localhost:3000
Vary: Origin
Access-Control-Max-Age: 60
Access-Control-Allow-Methods: OPTIONS, GET, POST, PUT, PATCH, DELETE
Access-Control-Allow-Headers: X-Requested-With, Content-Type, Accept, Origin, Authorization

The Access-Control-Allow-Origin echoes the Origin you sent. (The exact order of the methods and headers lists may differ between runs; treat them as sets.)

2. The actual request. This is the real call the browser makes once the preflight passes:

curl -i http://localhost:8081/your/path \
  -H 'Origin: http://localhost:3000'

The response carries the reflected origin and Vary alongside your stub's body:

HTTP/1.1 200 OK
Access-Control-Allow-Origin: http://localhost:3000
Vary: Origin
… (your stub's body and other headers) …

Drop the -H 'Origin: …' and you will see only Vary: Origin with no Access-Control-Allow-Origin; the reflection is keyed on the Origin a browser sends.

Reflecting any Origin is what makes mocks work for any browser client without configuration. If you would rather return a fixed Access-Control-Allow-Origin (a single allowed origin, or the wildcard *), set Access-Control-Allow-Origin yourself in the stub's response headers. When a stub supplies its own Access-Control-Allow-Origin, that value wins: the response carries exactly one Access-Control-Allow-Origin (your stub's value), and the auto-reflected one is suppressed, so the response stays CORS-compliant.

In a mapping JSON, set it under response.headers:

{
  "request": {
    "method": "GET",
    "url": "/your/path"
  },
  "response": {
    "status": 200,
    "headers": {
      "Access-Control-Allow-Origin": "https://app.example.org"
    },
    "body": "..."
  }
}

A request from any browser now receives Access-Control-Allow-Origin: https://app.example.org rather than its own reflected origin. Use "*" as the value to allow all origins explicitly.

You can set the same header from the GUI: open the mapping on the Add/Edit mappings page and add Access-Control-Allow-Origin in the response Headers field (the same field you use for any other response header). It serialises to the response.headers object shown above.

The trafficparrot.http.optionsResponse.enabled property (described under Extension System Properties) is a separate mechanism. It controls an automatic HTTP OPTIONS response that emits the standard Allow header (the list of methods a resource supports); it does not emit the Access-Control-* headers a browser CORS preflight needs. The out-of-the-box CORS behaviour described on this page is not provided by, and does not depend on, that property: your stub responses are CORS-friendly regardless of how optionsResponse.enabled is set.

By default the virtual service reflects any Origin. This is appropriate for a test double serving mock data (it lets any browser client call your mocks without configuration), but it is permissive. If you want to restrict which origins are allowed, set an explicit Access-Control-Allow-Origin (a specific origin) on the stub response, as shown in Pinning a specific origin above.

• Request matching: how Traffic Parrot picks a stub for an incoming request
• Dynamic responses: vary a stub response (including its headers) with the incoming request
• Add/Edit HTTP mappings: set response headers such as Access-Control-Allow-Origin in the GUI
• trafficparrot.properties reference: the optionsResponse.enabled property and other HTTP options

Use a single placeholder image for all image requests. This is ideal for basic testing where the actual image content doesn't matter (e.g., http://localhost:8080/images/any-skirt.jpg).

{
  "id": "a1b2c3d4-e5f6-7890-abcd-ef1234567890",
  "name": "Generic Image Response",
  "request": {
    "method": "GET",
    "urlPattern": "/images/.*"
  },
  "response": {
    "status": 200,
    "bodyFileName": "sample-image.jpg",
    "headers": {
      "Content-Type": "image/jpeg",
      "Cache-Control": "public, max-age=3600"
    }
  },
  "uuid": "a1b2c3d4-e5f6-7890-abcd-ef1234567890",
  "priority": 5
}

Serve different images based on URL patterns, useful when your application expects images of specific dimensions. You can create these using tools like ImageMagick on a Mac or any image editor (e.g., http://localhost:8080/images/320x240-skirt.jpg).

{
  "id": "b2c3d4e5-f6a7-8901-cdef-012345678902",
  "name": "320x240 Image Response",
  "request": {
    "method": "GET",
    "urlPattern": "/images/320x240.*"
  },
  "response": {
    "status": 200,
    "bodyFileName": "sample-image-320x240.jpg",
    "headers": {
      "Content-Type": "image/jpeg",
      "Cache-Control": "public, max-age=3600"
    }
  },
  "uuid": "b2c3d4e5-f6a7-8901-cdef-012345678902",
  "priority": 1
}

{
  "id": "b1c2d3e4-f5a6-7890-bcde-f12345678901",
  "name": "640x480 Image Response",
  "request": {
    "method": "GET",
    "urlPattern": "/images/640x480.*"
  },
  "response": {
    "status": 200,
    "bodyFileName": "sample-image-640x480.jpg",
    "headers": {
      "Content-Type": "image/jpeg",
      "Cache-Control": "public, max-age=3600"
    }
  },
  "uuid": "b1c2d3e4-f5a6-7890-bcde-f12345678901",
  "priority": 1
}

Use actual images from your application, either recorded during a session or manually configured. This provides the most accurate simulation (e.g., http://localhost:8080/images/product-123.jpg).

{
  "id": "c1d2e3f4-a5b6-7890-def1-123456789012",
  "name": "Product 123 Image Response",
  "request": {
    "method": "GET",
    "url": "/images/product-123.jpg"
  },
  "response": {
    "status": 200,
    "bodyFileName": "images/product-123.jpg",
    "headers": {
      "Content-Type": "image/jpeg",
      "Cache-Control": "public, max-age=86400",
      "ETag": "\"3e25960a79dbc69b674cd4ec67a72c62\""
    }
  },
  "uuid": "c1d2e3f4-a5b6-7890-def1-123456789012",
  "priority": 1
}

• Store image files in the __files directory or __files/images
• Use bodyFileName to reference the files
• Set appropriate Content-Type headers (image/jpeg, image/png, etc.)
• Consider adding Cache-Control headers for better performance
• Use priorities when combining multiple approaches

The HTTP passthrough proxy lets you forward requests that do not match any stub mapping to a real backend service. This is useful during development when you want to selectively mock only certain endpoints while letting all other traffic pass through to the real system.

When passthrough is enabled, Traffic Parrot automatically creates a low-priority catch-all proxy mapping on startup. Any request that does not match a higher-priority stub is forwarded to the configured target URL, and the real response is returned to the caller.

You can then inspect proxied requests in the Request Log and create stubs from them with a single click, allowing you to incrementally build up your virtual service.

To enable passthrough proxy mode, set the trafficparrot.virtualservice.proxy.defaultTargetUrl property to the base URL of the real backend service. You can set this in trafficparrot.properties or as a system property (e.g. via -Dtrafficparrot.virtualservice.proxy.defaultTargetUrl=http://backend:8080).

For example:

trafficparrot.virtualservice.proxy.defaultTargetUrl=http://my-backend:8080

On startup, Traffic Parrot creates a catch-all proxy mapping named "Default proxy to http://my-backend:8080" with priority 10. Because user-created stubs use default priority 5 (which is higher), any stub you create will automatically take precedence over the proxy.

The auto-created proxy mapping is visible in the HTTP Add/Edit mappings list. It cannot be edited or deleted from the UI because it is managed by the property. An info button appears in the actions column explaining the configuration.

To change the target URL, update the property value and restart Traffic Parrot. To remove passthrough entirely, clear the property value and restart.

To temporarily disable passthrough without removing the target URL configuration, set the trafficparrot.virtualservice.proxy.enabled property to false:

trafficparrot.virtualservice.proxy.enabled=false

When disabled, the auto-proxy mapping is not created on startup even if defaultTargetUrl is set. Requests that do not match any stub will return an unmatched response (status 900) instead of being forwarded.

The default value is true, so passthrough is enabled whenever defaultTargetUrl is configured.

When passthrough proxy mode is active, proxied requests appear in the Request Log with a blue "Proxied" badge in the Matched column instead of the green "Yes" badge. Hovering over the badge shows a tooltip with the proxy target URL (e.g. "Proxied to http://my-backend:8080").

This visual distinction makes it easy to see which requests were handled by stub mappings and which were forwarded to the real backend.

Each proxied request in the request log has a "Create Stub" button in its expanded detail view. Clicking it creates a new stub mapping based on the captured request and the actual response from the backend.

The created stub uses:

• The request HTTP method and an exact URL match (urlEqualTo)
• An exact request body match (when a request body is present)
• The actual response status code, headers, and body from the backend

Transport-specific headers such as Host, Connection, Transfer-Encoding, and Content-Length are automatically excluded from the stub because they are managed by the HTTP transport layer.

After the stub is created, a green notification appears with the message "Stub created for GET /your-path" (showing the actual method and URL). The notification includes a "View in Add/Edit" link that takes you to the HTTP Add/Edit page and opens the mapping editor for the new stub, where you can review and adjust the stub before use.

Once the stub is saved, subsequent requests to the same URL will be served by the static stub instead of being proxied to the backend. This allows you to incrementally build up your virtual service by exercising your application normally and converting real responses into stubs one at a time.

The "Create Stub" button only appears for proxied requests. It is not shown for requests that matched an existing stub or for unmatched requests.

Coming from WireMock's verify(...) DSL? In a WireMock JUnit test you assert that your code under test called a stub by writing, in the same JVM, something like verify(getRequestedFor(urlEqualTo("/x"))) or verify(exactly(0), getRequestedFor(urlEqualTo("/x"))). Traffic Parrot is a standalone virtual service, not an embedded library; there is no in-JVM WireMock.verify() client API to call. Instead you assert against the running server's request journal, the same __admin/requests journal WireMock's client calls under the hood.

You can do this two ways, and they read from the same journal:

• The Request Log GUI page (menu HTTP → Requests, served at /http/requests.html): browse every received request, expand a row to read its headers and body, and see why an unmatched request did not match. This is the interactive, point-and-click equivalent of findAll(...) and the near-miss report.
• The admin REST API under /api/http/…: copy-paste curl recipes you can run from a test, a CI step, or a terminal to assert a request was made, was made exactly N times, or was NOT made. This is the scriptable equivalent of verify(...) and getAllServeEvents().

The admin API is reachable on the Traffic Parrot management port (trafficparrot.gui.http.port, 8080 by default) under the /api/http/… prefix, which the GUI proxies through to the virtual service's /__admin/… journal. The recipes below all use the /api/http/… form. This section reframes verification around the WireMock mental model and gives runnable recipes; for the full GUI walkthrough and screenshots, it cross-links the Request Log section below rather than repeating it.

Each WireMock JUnit-client call maps onto either an admin-API recipe or a place to look in the Request Log GUI:

Note: count and find are POST, never GET. The match criteria travel in the request body, so /api/http/requests/count and /api/http/requests/find are POST endpoints. A GET to /requests/count is interpreted as a lookup of a single request whose id is the literal string count and will not return a count.

The equivalent of verify(getRequestedFor(urlEqualTo("/x"))) is to count the journal entries that match a set of criteria and assert the count is at least one. Send a POST to /api/http/requests/count with the criteria in the body:

curl -X POST http://localhost:8080/api/http/requests/count \
  -H 'Content-Type: application/json' \
  -d '{"method":"GET","url":"/x"}'

The response is the number of journal entries that matched:

{ "count": 1, "requestJournalDisabled": false }

A count of one or more means the request was made; your test asserts on the count field (the response also carries a requestJournalDisabled flag, which is false whenever the request journal is enabled, the default). The criteria body is a WireMock request pattern, so as well as method and url you can match on urlPath, urlPattern, headers, queryParameters, and bodyPatterns: the same matchers you use in a mapping (see Request matching). An empty body {} counts every request in the journal.

In the GUI, open the Request Log and look for the row: a matching request shows a green Matched = Yes badge.

The Request Log above is a live verification result: the count badge reads 3 requests, the two GET /example rows are Matched = Yes, and the unmatched POST /orders row is Matched = No. The Status column shows the status recorded in the journal: the unmatched row reads the configured no-match code 900 that the client received (see near misses below), and a matched row reads the status in the mapping's response definition.

This is the direct answer to the most common WireMock verification question: "verify that a request of a given URL has NOT been made", the equivalent of verify(exactly(0), getRequestedFor(urlEqualTo("/x"))). Use the same POST /api/http/requests/count call and assert the count is zero:

curl -X POST http://localhost:8080/api/http/requests/count \
  -H 'Content-Type: application/json' \
  -d '{"method":"GET","url":"/x"}'

If the request was never received, the count is zero:

{ "count": 0, "requestJournalDisabled": false }

A test can therefore assert "this endpoint was not called" by checking that count equals 0. In the GUI, the same conclusion is that no row matching that method and URL appears in the Request Log.

The equivalent of verify(exactly(3), getRequestedFor(...)) is the same POST /api/http/requests/count call; you just assert the exact number returned:

curl -X POST http://localhost:8080/api/http/requests/count \
  -H 'Content-Type: application/json' \
  -d '{"method":"GET","url":"/x"}'

{ "count": 3, "requestJournalDisabled": false }

There is one endpoint for "at least once", "exactly N", and "never": the assertion lives in your test (count >= 1, count == 3, or count == 0), not in a different URL.

When you need more than a count (the equivalent of findAll(getRequestedFor(...)) followed by reading request.getBodyAsString()), use find to return the matching requests in full, including their bodies and headers. Send a POST to /api/http/requests/find with a request-matcher body:

curl -X POST http://localhost:8080/api/http/requests/find \
  -H 'Content-Type: application/json' \
  -d '{"method":"POST","url":"/orders"}'

The response lists each matching request with its method, URL, headers, and captured body, so you can assert on the exact payload your code under test sent. To pull back the whole journal (the equivalent of getAllServeEvents()) use:

curl http://localhost:8080/api/http/requests

{ "requests": [ ... ], "meta": { "total": 7 }, "requestJournalDisabled": false }

In the GUI, click the expand arrow on a row to read the request headers and body (JSON and XML bodies are pretty-printed, with a Copy button for the original bytes); see Viewing request details and Pretty-printed bodies.

When the journal holds a lot of requests you can narrow what GET /api/http/requests returns with two optional query parameters. Both are off by default: a plain GET /api/http/requests with neither parameter returns the whole journal exactly as before, so existing scripts and integrations are unaffected.

• limit: a non-negative integer that bounds the response to the first N entries in journal order. limit=0 returns no rows; a negative value is rejected with HTTP 400.
• since: an ISO-8601 timestamp (for example 2026-07-16T10:00:00Z). Only requests recorded strictly after that instant are returned (the boundary instant itself is excluded), compared against each entry's recorded time. A value that is not a valid ISO-8601 timestamp is rejected with HTTP 400.

For example, to fetch only the first two entries in journal order, or only the requests received after a known point in time:

curl 'http://localhost:8080/api/http/requests?limit=2'
curl 'http://localhost:8080/api/http/requests?since=2026-07-16T10:00:00Z'

limit and since can be combined, and either may be used on its own. The meta.total field always reports the full number of entries in the journal before limit or since is applied (it is deliberately independent of both parameters), so it stays a reliable count of everything recorded even when the requests array has been trimmed:

{ "requests": [ ...at most 2 rows... ], "meta": { "total": 7 }, "requestJournalDisabled": false }

If a request did not match any mapping, Traffic Parrot serves an HTTP 900 response ("No responses matched") rather than throwing an exception, and records the request as unmatched in the journal. To list the unmatched requests, the equivalent of findUnmatchedRequests():

curl http://localhost:8080/api/http/requests/unmatched

To find out why each unmatched request missed (the equivalent of findNearMissesForAllUnmatched()), ask for the near misses, which report the closest mappings and a per-field comparison of expected vs actual:

curl http://localhost:8080/api/http/requests/unmatched/near-misses

The GUI presents the same near-miss diagnosis interactively: expand an unmatched row in the Request Log to see a field-by-field comparison table (Method, URL, Headers, Body) with match/mismatch indicators, the closest stub's name and match percentage, and a link to edit that stub; see Near misses for unmatched requests.

So that each test verifies only the requests it made, clear the journal first, the equivalent of resetRequests(). Send a DELETE to /api/http/requests:

curl -X DELETE http://localhost:8080/api/http/requests

In the GUI, the red Clear button on the Request Log does the same thing (with a confirmation prompt); see Clearing the request log.

The verification endpoints, all on the management port (8080 by default) under the /api/http/… prefix:

These are reachable through the management port's /api/http/… write-through prefix shown above. The corresponding raw journal routes on the virtual service port are /__admin/requests and friends; prefer the /api/http/… form, which is the supported management surface. For an interactive catalogue of the management APIs, see Management APIs in the user guide (Postman workspace and OpenAPI specification).

• Request Log: the GUI page that shows every received request, with screenshots of the table, expandable details, and near-miss diagnosis
• Request matching: the matchers (URL, headers, query parameters, body patterns) you can put in a count or find criteria body, the same ones you use in a mapping
• Management APIs: the Postman workspace and OpenAPI specification for the full set of Traffic Parrot management endpoints

The Request Log page shows all HTTP requests received by the virtual service. You can access it from the HTTP dropdown menu in the top navigation bar by clicking "Requests".

Each row in the table displays the request timestamp, HTTP method, URL, response status code, whether the request was matched to an existing stub mapping, and the scenario the request was served under. Matched requests show a green "Yes" badge, unmatched requests show a red "No" badge, and proxied requests show a blue "Proxied" badge (see HTTP Passthrough Proxy).

The Scenario column shows the name of the scenario the request was served under. Requests served under the default scenario are shown as "Default".

The page header shows the total number of requests and a "Last updated" timestamp indicating when the data was last refreshed.

Click the expand arrow on any row to view full details of the request and its response. The expanded section shows the request headers, request body, response status, response headers, and response body.

For matched requests, a "Matched Stub" section displays the full JSON definition of the stub mapping that was used to generate the response. This is useful for understanding which mapping handled the request.

When a request or response body is JSON or XML, it is automatically pretty-printed (indented and shown over multiple lines) in the expanded detail section, so that you can read the structure at a glance instead of scanning a single long line.

Each body block is collapsible: click the heading ("Request Body" or "Response Body") to fold it away when you want to focus on the rest of the detail panel. A Copy button next to each body copies the original, unformatted body to your clipboard, so what you paste matches the bytes that were actually sent or received, not the pretty-printed version shown on screen.

The same pretty-printing and Copy button are available wherever request bodies are logged, including the gRPC requests page and the JMS, IBM MQ, and file messaging requests pages.

Bodies are shown exactly as received (raw, on a single line) when they cannot be safely formatted, for example when the body is not well-formed JSON or XML, is not a JSON or XML content type, or is large enough to be truncated for display (bodies over 10 KB are truncated, with a note showing the total size).

When you expand an unmatched request (one with a red "No" badge), the detail section shows a "Near Misses" list instead of a matched stub. Near misses are the existing stub mappings that came closest to matching the request, along with a percentage indicating how close each match was.

Each near miss displays a per-field comparison table that shows exactly which fields matched and which did not. The table compares the stub's expected values against the actual request values for each field: Method, URL, Headers, and Body.

Each row in the comparison table has an indicator on the right:

• A green checkmark means the field matched the stub's expected value
• A red cross means the field did not match; mismatched rows are highlighted with a red background so you can quickly spot the problem
• A grey question mark means the match could not be evaluated in the browser (for example, JSONPath, XPath, or XML body patterns)

In the example above, a request to GET /api/users-wrong with an Accept: text/html header is compared against a stub expecting GET /api/users with Accept: application/json. The Method matches (green checkmark), but both the URL and Accept header are mismatched (red crosses), making it immediately clear why the request was not matched.

Each near miss entry shows the stub's human-readable name as its heading, when one has been set. If the stub does not have a name, the heading displays the stub's method and URL pattern instead. Next to the name, an orange badge shows the match percentage (for example, "92% match"), which indicates how closely the request matched that stub.

Each near miss entry includes a "View Stub" link next to the stub name. Clicking this link navigates directly to the HTTP Stubs page and opens the edit dialog for that specific stub mapping. This allows you to quickly fix the stub's request pattern when you can see which field did not match.

Below the comparison table, a "Show full stub JSON" link is available for users who need to see the complete stub mapping definition. Clicking this link toggles a collapsible section that displays the raw JSON of the stub, including the response body, metadata, and UUID. This is hidden by default to keep the near miss display focused on the comparison table.

The request log automatically refreshes every second, so new requests appear in the table without requiring a manual page reload. This is useful when monitoring live traffic to the virtual service.

You can pause auto-refresh by clicking the "Stop auto reload" button. When paused, the button changes to "Start auto reload" and no further automatic updates occur until you resume. The manual "Refresh" button continues to work at any time.

Auto-refresh also pauses automatically when you expand a detail row, select text in the table, or hover over a tooltip. This prevents the data from changing while you are inspecting it.

Click the red "Clear" button to delete all recorded requests from the journal. A confirmation dialog appears asking you to confirm before the requests are permanently removed.