yamantaka520/Agent-Memory-OS

Overview

Local-first memory engine for AI-agent teams. Provides private, team, and project access control, associative recall, federated sync, and a built-in MCP server. Stores everything in one SQLite file and does not require an LLM. Apache-2.0 licensed.

Best for

Best for
Teams building local-first AI agents that need shared memory without cloud dependencies

Use cases

• Share agent memory across a team with access control
• Sync memory between local and remote agent environments
• Expose memory to agents through the built-in MCP server

How to use

Install

pip install 'agent-memory-os[full]'    # recommended: everything (Web UI, MCP, turbovec)

Tested with

Claude Code, ChatGPT

Notes

Local-first memory engine for AI-agent teams. Provides private, team, and project access control, associative recall, federated sync, and a built-in MCP server. Stores everything in one SQLite file and does not require an LLM. Apache-2.0 licensed.

5 stars on GitHub. Last updated 2026-09-02. Licensed Apache-2.0.

Use cases

• Share agent memory across a team with access control
• Sync memory between local and remote agent environments
• Expose memory to agents through the built-in MCP server

Pros

• Single SQLite file simplifies deployment and portability
• No LLM dependency reduces cost and latency
• Apache-2.0 license permits flexible reuse

Cons

• Federated sync adds setup overhead for multi-device use
• Associative recall is not semantic search without an LLM
• Single-file storage limits horizontal scaling

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.