Base URL and authentication
All calls go tohttps://api.mixpeek.com and carry an API key as a bearer token:
401. A key without the permission an operation needs returns 403; see API keys for permissions and scopes.
The X-Namespace header
Buckets, collections, retrievers, taxonomies, clusters and documents live inside a namespace. Calls on them need the namespace id in X-Namespace:
ns_...) that POST /v1/namespaces returned, not the namespace name. Organization-level calls, such as storage connections, webhooks and API keys, take no X-Namespace. Each endpoint page lists the headers it needs.
Response envelopes
A retriever run that times out or skips a stage still answers
200, with status: "degraded" and an entry in warnings. Check status before you treat a short documents list as “no matches”. A clean run returns status: "completed".
Errors
Every error has the same body:error.type, and on error.code where an endpoint sets one, such as namespace_name_taken. message is for people and can change. A 429 carries Retry-After in seconds; see Rate limits and quotas.
Pagination
List endpoints returnnext_cursor. Send it back as cursor and stop when the response omits it. next_page and previous_page are always null. See Pagination.
Machine-readable docs
- OpenAPI spec:
https://api.mixpeek.com/docs/openapi.json, the spec the API serves now. - Docs index for agents:
https://mixpeek.com/docs/llms.txt, and the full text at/docs/llms-full.txt. - Any docs page as Markdown: add
.mdto its URL, for example/docs/overview/quickstart.md. - MCP: connect an agent to the API directly with the MCP server.

