Store BEARQ_API_KEY as a secret in your CI runner. The script below runs a full test suite, streams all tasks in parallel over SSE, and exits non-zero if any test run fails.

IDS=$(curl -sf -X POST "$BEARQ_BASE_URL/tasks" \
  -H "Authorization: Bearer $BEARQ_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"agent":"tester","mode":"run"}' | jq -r '.taskIds[]')

run_task() {
  curl -sN -H "Authorization: Bearer $BEARQ_API_KEY" \
    "$BEARQ_BASE_URL/tasks/$1/stream" \
  | awk -v id="$1" '
    /^event: done$/ { getline; sub(/^data: /, ""); print id, $0; exit }
  ' \
  | jq -e --arg id "$1" '.testRunStatus == "passed"' >/dev/null
}

PIDS=()
for ID in $IDS; do run_task "$ID" & PIDS+=($!); done

FAILED=0
for PID in "${PIDS[@]}"; do wait "$PID" || FAILED=1; done
exit $FAILED

{
  "type": "https://problems-registry.smartbear.com/unauthorized",
  "title": "Unauthorized",
  "status": 401,
  "detail": "Invalid API key",
  "code": "401-01"
}

Creates one or more tasks and assigns them to a BearQ agent. The agent field discriminates between three variants. A single request may produce multiple tasks — for example, a Tester request scoped to a functional area creates one task per test case in that area.

On success, poll GET /tasks/{taskId}/status for each returned ID, or open GET /tasks/{taskId}/stream to be notified of completion without polling.

{ "agent": "qa-lead", "instruction": "Add important new test cases for our product page." }

{
  "agent": "tester",
  "mode": "run",
  "functionalAreas": ["Checkout"]
}

{
  "agent": "tester",
  "mode": "refine",
  "testCaseIds": [4821, 4822, 4823]
}

{ "agent": "tester", "mode": "run" }

{ "agent": "explorer", "functionalArea": "Checkout" }

{ "taskIds": [9183, 9184, 9185] }

Returns full metadata for a task along with its complete activity log. Use this for detailed inspection — for example, displaying progress in a UI or surfacing activity messages to a developer.

For lightweight polling, prefer GET /tasks/{taskId}/status. To follow a task in real time, use GET /tasks/{taskId}/stream.

{
  "taskId": 9183,
  "taskStatus": "completed",
  "metadata": {
    "type": "tester",
    "name": "Checkout",
    "parentTaskId": null,
    "subtaskIds": [],
    "userId": 412,
    "createdAt": "2026-05-15T18:04:21.123Z",
    "completedAt": "2026-05-15T18:07:55.901Z",
    "duration": "PT3M34.778S",
    "acus": 0.14,
    "testCaseId": 4821,
    "testRunStatus": "passed"
  },
  "activityLog": [
    {
      "messageType": "status-message",
      "timestampAt": "2026-05-15T18:04:21.200Z",
      "description": "Started"
    },
    {
      "messageType": "action-message",
      "timestampAt": "2026-05-15T18:04:22.501Z",
      "description": "Navigated to /cart"
    },
    {
      "messageType": "action-message",
      "timestampAt": "2026-05-15T18:04:24.118Z",
      "description": "Applied coupon SAVE10"
    }
  ]
}

Cancels a task that is pending or running. Tasks that have already reached a terminal state (completed, failed, or canceled) are unaffected — the request still returns 200.

Calling this on an already-terminal or already-deleted task returns 200 with an empty object body.

Returns only the task's current lifecycle status. This is the lightweight endpoint for polling. To avoid polling entirely, open GET /tasks/{taskId}/stream and wait for the done event.

Recommended CI/CD pattern: Create tasks with POST /tasks, then either poll this endpoint for each taskId until the status is terminal (completed, failed, or canceled), or stream each task via SSE for a push-based alternative.

{ "status": "running" }

Streams a task's metadata and activity log as Server-Sent Events (SSE). Two primary patterns:

The stream works for both running and already-terminal tasks. If the task is already terminal when the stream opens, all events are flushed immediately, and the stream closes.

For one-shot reads, prefer GET /tasks/{taskId}. For lightweight polling, prefer GET /tasks/{taskId}/status.

event: metadata
data: {"taskId":9183,"taskStatus":"running","metadata":{"type":"tester",
  "name":"Checkout","parentTaskId":null,
  "subtaskIds":[],"userId":412,"createdAt":"2026-05-15T18:04:21.123Z",
  "completedAt":null,"duration":null,
  "testCaseId":4821,"testRunStatus":"running"}}

event: activityLogEntries
data: [{"messageType":"action-message","timestampAt":"2026-05-15T18:04:22.501Z",
  "description":"Navigated to /cart"}]

event: activityLogEntries
data: [{"messageType":"action-message","timestampAt":"2026-05-15T18:04:24.118Z",
  "description":"Applied coupon SAVE10"}]

event: done
data: {"status":"completed","result":"All assertions passed.",
  "testCaseId":4821,"testRunStatus":"passed"}

Liveness probe used by infrastructure to verify the public API is running. Returns the plain-text string OK with status 200. This endpoint does not require authentication and is not intended for application use.

A task moves through these lifecycle states. Once it reaches a terminal state, it will not change again.

Returned inside TaskDetailsResponse and in the metadata SSE event payload.

Outcome enum for Tester tasks only. Distinct from the task's lifecycle TaskStatus.

Each entry in activityLog (and in activityLogEntries SSE frames) has a messageType field. The following types are unstable and may change without notice: