Get started with MemTap

Deploy graph-based long-term memory for your AI agents in under 5 minutes. Self-hosted, open source, no cloud dependency.

1

Clone and configure

Clone the repository and set up your environment file. The defaults work for local development.

$ git clone https://github.com/psifactory/memtap.git $ cd memtap $ cp .env.example .env # Edit .env to set your LLM API key (optional, for auto-capture)
2

Start the server

One command brings up MemTap and ArangoDB. Both run in Docker containers.

$ docker compose up -d # MemTap server: http://localhost:18800 # ArangoDB UI: http://localhost:8529 # Health check: http://localhost:18800/health
3

Connect your agent

Add MemTap to your agent configuration. Native OpenClaw plugin shown below, or use the REST API with any framework.

// openclaw.json { "plugins": { "memtap": { "serverUrl": "http://localhost:18800", "agentId": "my-agent", "autoCapture": true, "bulletinOnBoot": true } } }
4

Store your first memory

Test the API by storing a memory. Entities are auto-extracted and linked in the knowledge graph.

$ curl -X POST http://localhost:18800/memories \ -H "Content-Type: application/json" \ -d '{ "type": "fact", "content": "Our project uses TypeScript and PostgreSQL", "agent": "my-agent", "importance": 0.8 }' # Response: memory stored, entities extracted, edges created
5

Recall memories

Search across all memory types with BM25 ranking. Results include graph context.

$ curl "http://localhost:18800/recall?q=database&agent=my-agent" # Returns ranked memories with entity links and edge counts

The free tier includes 500 memories, BM25 search, entity linking, and decision tracking.
Upgrade to Pro for GraphRAG, memory decay, contradiction detection, and more.

View Pricing API Documentation