Character QA API

Evaluate your agent’s persona consistency with scoring, drift detection, and configurable probe sets. These endpoints proxy requests to the agent’s OpenClaw runtime.

All character QA endpoints require an authenticated session. The proxy resolves your agent’s URL from the database and forwards requests to the running instance. If no agent is deployed, the endpoint returns a 404 with status: "no_agent".

Get evaluation results

GET /api/openclaw/character-qa

Returns the agent’s character evaluation history and current persona scores. Proxies to the agent’s GET /api/eval/character endpoint.

Response

{
  "scores": {
    "voice": 0.92,
    "emotion": 0.87,
    "knowledge": 0.95,
    "refusal": 0.88
  },
  "driftDetected": false,
  "history": [
    {
      "id": "eval_001",
      "timestamp": "2026-04-09T10:00:00Z",
      "probeSet": "default",
      "dimensions": ["voice", "emotion", "knowledge", "refusal"],
      "overallScore": 0.905,
      "driftDetected": false
    }
  ]
}

Field	Type	Description
`scores`	object	Current persona dimension scores (0–1 scale)
`scores.voice`	number	How well the agent maintains its configured voice and tone
`scores.emotion`	number	Emotional consistency and appropriate emotional responses
`scores.knowledge`	number	Accuracy and consistency of knowledge domain responses
`scores.refusal`	number	Appropriate handling of out-of-scope or harmful requests
`driftDetected`	boolean	`true` when the agent’s persona has drifted significantly from its baseline
`history`	array	List of past evaluation runs

Errors

Code	Description
401	`Unauthorized` — no authenticated session
404	`No agent deployed` — the user has no running agent instance
502	`Agent unreachable` — the agent instance did not respond within 10 seconds

Run a character evaluation

POST /api/openclaw/character-qa

Triggers a new persona evaluation run. Proxies to the agent’s POST /api/eval/character/run endpoint.

Request body

Field	Type	Required	Description
`probeSet`	string	No	The probe set to use for evaluation. Defaults to `"default"`.
`dimensions`	string[]	No	Which persona dimensions to evaluate. Defaults to `["voice", "emotion", "knowledge", "refusal"]`.

Example

curl -X POST https://agentbot.sh/api/openclaw/character-qa \
  -H "Content-Type: application/json" \
  -d '{
    "probeSet": "default",
    "dimensions": ["voice", "emotion", "knowledge", "refusal"]
  }'

Response

The response contains the evaluation results from the agent, matching the structure returned by GET /api/openclaw/character-qa with the addition of the newly completed run.

Errors

Code	Description
400	`Invalid request body` — the request body is not valid JSON
401	`Unauthorized` — no authenticated session
404	`No agent deployed` — the user has no running agent instance
502	`Agent unreachable` — the agent instance did not respond within 15 seconds

Getting Started

Core Concepts

Integrations

Core Services

Payments

API Reference

MCP

Security

Character QA API

Character QA API

Get evaluation results

Response

Errors

Run a character evaluation

Request body

Example

Response

Errors

​Character QA API

​Get evaluation results

​Response

​Errors

​Run a character evaluation

​Request body

​Example

​Response

​Errors

Character QA API

Get evaluation results

Response

Errors

Run a character evaluation

Request body

Example

Response

Errors