Documentation Index
Fetch the complete documentation index at: https://docs.openmem.blog/llms.txt
Use this file to discover all available pages before exploring further.
omp-server is a FastAPI application that exposes the full OMP verb set as HTTP endpoints. Because it speaks the OMP protocol natively, any language or framework — not just Python — can store, search, and retrieve memories by making ordinary HTTP requests. You can also connect back to it from the Python SDK using the passthrough provider, which gives you a single server process shared across multiple clients or services.
Install
Start the server
Available routes
| Method | Path | Description |
|---|---|---|
POST | /memories | Add a memory |
GET | /memories | List memories |
GET | /memories/{id} | Get a memory by ID |
PATCH | /memories/{id} | Update a memory |
DELETE | /memories/{id} | Delete a memory |
GET | /memories/search | Semantic and keyword search |
POST | /context | Get a prompt-ready context block |
GET | /audit | Retrieve the audit log |
GET | /capabilities | Provider capabilities |
GET | /healthz | Health check |
Connect using the Python SDK
Point the Python SDK at the running server using thepassthrough provider. All standard Memory methods work exactly as they do against a direct adapter.
/capabilities on construction. If the server returns omp_version, the SDK routes all calls through it as a native OMP server.
Check server health
200 unconditionally to avoid unnecessary calls to paid external endpoints.
Server behavior
CORS is disabled by default. If you need browser clients to call the server directly, enable it with the
--cors-origins flag:The default request body limit is 1 MiB. Requests larger than this limit receive a
413 response with code: payload_too_large before any schema validation runs. Increase the limit with --max-request-bytes if your use case requires larger payloads.