Tell us what you think.

Was an answer reliable or unreliable? Did the evidence look trustworthy? Was something confusing? Pick one. The two text boxes are optional.

For agents

Send one rating after you act on an answer. No key. Include the decision_id when the feedback is about a decision; every decision names this endpoint in its feedback block.

• Endpoint: POST https://api.modelspec.dev/v1/feedback
• Request schema: https://modelspec.dev/api/feedback/v1.schema.json
• Ratings: reliable, unreliable, trustworthy, untrustworthy, confusing. Client: agent, cli, mcp or page.
• CLI: modelspec feedback DECISION_ID --rating unreliable --note "…"
• MCP: the feedback tool on https://api.modelspec.dev/mcp
• Reference: docs/feedback-api.md and operation feedback in openapi.yaml

curl -X POST https://api.modelspec.dev/v1/feedback -H 'content-type: application/json' -d '{"rating": "unreliable", "client": "agent", "decision_id": "dec_…", "note": "What went wrong"}'

What we keep

Feedback storage is switched off. Until it is switched on, what you send is checked and answered, and nothing of it is kept. Before it is switched on, the privacy statement will say exactly what is kept, for how long, and how to delete it.

Whatever happens, feedback needs no account and no key, and your address is never stored with it. Please don't include prompts, keys or personal details in the text boxes.

What you told us, and what we changed

Nothing yet. Feedback storage is not switched on, so there is no feedback to act on. Each change made because of what you tell us will be listed here, with its date and what shipped.