Mnemexa API

The Intelligent Memory OS for AI. Hosted memory infrastructure that stores, retrieves, and self-optimizes — accessed through a four-endpoint REST API or the official Python SDK and MCP adapter.

API v1 · Productionhttps://api.mnemexa.com

What you get

Rather than managing vector databases, embedding pipelines, and ranking heuristics, you make four HTTP calls and get back semantically-ranked, importance-weighted memories. Every stored memory is automatically:

PII-screened — passwords, credit cards, SSNs, and API keys are detected and rejected before they reach the database.
Semantically deduplicated — near-identical memories are merged, preserving the richer version.
Importance-scored — an LLM rates business value so high-signal memories rank above noise.
Temporally classified — persistent facts (“client prefers async”) are distinguished from time-bound state (“meeting at 3pm tomorrow”).
Categorized — multi-tag labels for downstream filtering.

Retrieval combines cosine similarity with recency, importance, and access frequency — see Hybrid Scoring.

The four endpoints

Endpoint	Purpose	Page
`POST /v1/memory/store`	Save a memory through the intelligence pipeline.	memory.store
`POST /v1/memory/retrieve`	Semantic retrieval with hybrid scoring.	memory.retrieve
`GET /v1/optimize/health`	Workspace memory health diagnostic.	optimize.health
`GET /v1/status`	Connection check and workspace identity.	status

The Python SDK and MCP Adapter wrap these four endpoints with idiomatic interfaces — the underlying surface is identical.

Base URL

https://api.mnemexa.com

All requests must use HTTPS. Request and response bodies are JSON (Content-Type: application/json). Timestamps are ISO 8601 UTC.

Authentication

Every request requires a workspace API key in the Authorization header:

Authorization: Bearer mnx_ws_…

See Authentication for how to provision keys.

Three ways to call the API

For most AI-agent integrations, install the MCP adapter — your agent gets memory automatically with no code changes. For backend services, use the Python SDK. For everything else, the REST API is what both adapters wrap.

# 1. REST — works from anything
curl -X POST https://api.mnemexa.com/v1/memory/store \
  -H "Authorization: Bearer mnx_ws_…" \
  -H "Content-Type: application/json" \
  -d '{"text": "Client prefers async standups."}'

# 2. Python SDK
import mnemexa
client = mnemexa.Client()
client.memory.store(text="Client prefers async standups.")

# 3. MCP — for AI agents (Claude, Cursor, Windsurf, etc.)
npx @mnemexa/mcp
# Then your agent calls brain.remember / brain.recall automatically.

Continue with the Quick Start to get a key and store your first memory in under five minutes.

Next Quick Start →