# agentic-memory MCP server

> Provides AI agents with a persistent cognitive graph memory to store facts, decisions, reasoning chains, and corrections in a single navigable file.

- Repository: https://github.com/agentralabs/agentic-memory
- Page: https://hostedmcp.org/servers/agentralabs/agentic-memory
- Transport: —
- Stars: 32
- License: MIT

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `AGENTIC_TOKEN` | no | no |  |

## Tools (13)

- `memory_capture_message`
- `memory_capture_tool`
- `memory_capture_file`
- `memory_capture_decision`
- `memory_capture_boundary`
- `memory_retrieve`
- `memory_resurrect`
- `memory_v3_session_resume`
- `memory_search_temporal`
- `memory_search_semantic`
- `memory_search_entity`
- `memory_v3_stats`
- `memory_verify_integrity`

## FAQ

### How does AgenticMemory solve the problem of AI amnesia?

It stores an agent's knowledge as a navigable graph in a single binary file, preserving facts, decisions, reasoning chains, and corrections across sessions.

### What kind of query types are supported?

The system supports 16 query types, including temporal, semantic, causal, entity, and procedural searches.

### Is the memory data secure and permanent?

Yes, the V3 architecture uses an append-only storage layer with BLAKE3 integrity chains, making the memory cryptographically tamper-proof.

### Which AI clients can use this memory?

It supports auto-syncing with Claude Code, Cursor, Windsurf, and Cody.