Tell us what would make 0xArchive more useful. You can use the feedback form, call POST /v1/feedback from an application, or let a connected agent call the hosted MCP submit_feedback tool. Reports go to a private support queue. A receipt confirms that we received the report; it does not promise a reply or a fix.
REST
Applications can call this endpoint without credentials or include a valid 0xArchive API key in X-API-Key to associate the report with an account. A supplied invalid key is rejected. Submitting feedback does not use market-data credits. Agents may decide when to send feedback without asking before every submission. Configure your agent to send concise, relevant reports rather than every error it encounters.
An accepted request returns HTTP 202 with success: true and a data object containing status: "received" and an id for your records. type is bug, feature, question, or other. To flag a security report, also set category to security; otherwise it defaults to support. message is required and limited to 5,000 bytes. Optional reproduction fields are top-level: surface, route, venue, market_family, symbol, data_type, start, end, request_id, and error_code. start and end are Unix milliseconds. For a data issue, include the route, venue, market family, symbol, time window, observed result, expected result, and request ID where available. Do not invent details you cannot verify. A deliberate retry can carry the same Idempotency-Key (8–128 visible ASCII characters (no spaces)); reusing a key with different content is rejected.
See the request and response schema for field-level types and bounds.
The public website form accepts feedback without an API key. If you are signed in, the browser session is validated by the server and the report is associated with your account. An invalid supplied credential is rejected rather than treated as anonymous feedback. Submission limits apply; back off when you receive HTTP 429 and use Retry-After when present.
Hosted MCP
The hosted MCP server uses OAuth, not an 0xArchive API key. To let an agent call submit_feedback, reconnect through a client that explicitly requests mcp:feedback.write and approve that scope once. A client that reconnects asking only for mcp:market.read remains read-only; an existing read-only connection cannot submit. After the scope is granted, the agent may choose to submit feedback without asking before each call. You can remove access by disconnecting the integration.
The tool returns the same receipt ID. It does not send feedback on every API error automatically.
What to include
- What happened or what you want to change, in your own words.
- For a reproducible problem: the route or tool, venue, market family, symbol, UTC time window, and
request_id if available.
- For a suggestion: who it would help and why the current workflow falls short.
Do not include API keys, OAuth tokens, wallet secrets, private keys, raw conversation transcripts, full datasets, or sensitive account data. Security reports use the same private queue, with ordinary support-team access; redact exploit secrets and credentials. We delete identifiable feedback and related delivery records from our feedback database, support inbox and CRM within three years. You can request earlier deletion at [email protected]. De-identified product insights may be kept longer. For an immediate account or service issue, contact support; feedback is not an urgent-response channel. Last modified on October 31, 2026