> ## Documentation Index
> Fetch the complete documentation index at: https://docs.allocait.app/llms.txt
> Use this file to discover all available pages before exploring further.

# Agent feedback

> How an agent reports a gap in the API, MCP server or docs.

Use agent feedback for a tool description that's ambiguous, an error you didn't expect from an endpoint, a missing capability, or docs that are wrong. It lets an agent calling Allocait report friction back, separate from anything about the user it's acting for. Allocait's instructions tell an agent to send it only when its user asks. Allocait attaches no portfolio data to it and stores only the fields in the request.

```bash theme={null}
curl -X POST "https://allocait.app/api/v1/agent-feedback" \
  -H "x-api-key: $ALLOCAIT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "kind": "docs",
    "summary": "symbols_to_price is not explained on the snapshot page",
    "tool_or_route": "GET /portfolio/snapshot"
  }'
```

This call needs an API key with the `write` permission. An account can file 20 per rolling 24 hours (a `429` with `DAILY_LIMIT_EXCEEDED` after that); see [Rate limits](https://docs.allocait.app/api/rate-limits.md).

`kind` is one of `bug`, `docs`, `missing_capability`, or `other`, and `summary` (up to 200 characters) is the only other required field. `details`, `tool_or_route`, `error_code`, `request_id`, and `client` (REST only) are all optional context that helps the team track down the issue faster.

If you send the same report again while the first is still open and recent, Allocait returns the first report instead of filing a second, and the repeat doesn't count toward the daily cap. Read your feedback back with `GET /agent-feedback`. `status` is `open`, `acknowledged` or `resolved`.

Through MCP, one tool, `send_feedback`, covers agent feedback, support requests and early access feedback (see [Send feedback](https://docs.allocait.app/guides/send-feedback.md)). Allocait's instructions tell the agent to call it only when you ask.

| `topic` | Use for | `kind` |
| - | - | - |
| `account` | Your account, billing or a product problem. Needs **Read and change**; on a read-only connection use **Contact support** in the app. `details` is required. | `bug`, `billing`, `feature_request`, `other` |
| `tools` | A tool, its description or the docs. | `bug`, `docs`, `missing_capability`, `other` |
| `early_access` | How a task went. | `bug`, `docs`, `missing_capability`, `works_well`, `other` |

`summary` (up to 200 characters) is required. `details` takes up to 1,000 characters. For `tools`, `tool`, `error_code` and `request_id` help the team find the problem. The tool takes no client field. Keep holdings and account data out of the text. The result is `{topic, filed, message}`; `early_access` also returns a `ref`. The message tells the agent not to promise a reply.

## Next steps

* [Send feedback](https://docs.allocait.app/guides/send-feedback.md): support requests and early access feedback.
* [Errors](https://docs.allocait.app/claude/errors.md): when an agent should file a `tools` report.
* [Action and feedback tools](https://docs.allocait.app/claude/tools/actions-feedback.md): the `send_feedback` fields.
* [Rate limits](https://docs.allocait.app/api/rate-limits.md): the daily cap.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.