This documentation reflects the current MVP version of the Flumes Memory API (path prefix /v1) implemented in the repository /flumes. Future versions may add additional endpoints and data stores. All examples in this section assume the default in-memory FAISS vector store unless otherwise noted.

​
Base URL

https://api.flumes.ai/v1
This endpoint is served from Flumes Cloud at https://api.flumes.ai. If you run the API locally the base URL is typically http://localhost:8000.

​
Authentication

All requests must include an API key using the Authorization header:
​
Authorization
string
required
Bearer token for authentication. Format: Authorization: Bearer <YOUR_API_KEY>.
You can generate and manage your API keys from the Flumes dashboard at app.flumes.ai.
If the key is missing or invalid the server responds with 401 Unauthorized.

​
Rate limits

Each API key is limited to 60 requests per minute by default. Exceeding this limit returns 429 Too Many Requests. The limit can be adjusted with the FLUMES_RATE_LIMIT environment variable on the server.

​
Error handling

All error responses share a common JSON structure: 
{
  "detail": "Human-readable error message"
}
The following table lists standard HTTP status codes returned by the API:
CodeMeaningTypical causes
400Bad RequestMissing required parameters, invalid JSON payload
401UnauthorizedAPI key missing / revoked
404Not FoundMemory id does not exist
429Too Many RequestsRate limit or monthly quota exceeded
500Internal Server ErrorUnhandled exception on the server

Add memory

Create new memories from chat messages.

Get memory

Retrieve a single memory by ID.

Search memories

Semantic search or list recent memories.

Update memory

Modify an existing memory.

Delete memory

Permanently remove a memory.