Repolode
Submit
rohitg00

rohitg00/agentmemory

Persistent memory for AI coding agents that captures sessions automatically and injects relevant context into later sessions across Claude Code, Cursor, Codex and other MCP clients.

View on GitHubWebsite
Stars
29,683
Forks
2,621
Open issues
582
Last commit
22 hours ago

What is it?

agentmemory is a memory server for AI coding agents. It silently records what an agent does during a session, compresses those observations into searchable memory, and feeds the relevant context back when the next session starts, so developers do not have to re-explain architecture, past bugs, or preferences each time.

It runs locally on top of the iii engine and SQLite, with no external database required. Retrieval combines BM25 keyword search, optional local vector embeddings (all-MiniLM-L6-v2), and graph matching. Without an embedding provider it runs keyless using BM25. It exposes 54 MCP tools and a REST API on port 3111, and includes a real-time viewer on port 3113 with session replay.

Integrations cover Claude Code, Codex CLI, GitHub Copilot CLI, Cursor, Gemini CLI, OpenClaw, Hermes, OpenCode, Devin and many other MCP-capable clients, and all of them share one memory server.

Key features

  • 01Automatically captures agent activity through lifecycle hooks, with no manual saving required.
  • 02Hybrid retrieval combining BM25, optional local vector embeddings and graph matching.
  • 03Exposes 54 MCP tools and a REST API for any MCP-compatible client.
  • 04Runs locally with SQLite and the iii engine, without external databases.
  • 05Includes a real-time viewer with session replay and JSONL transcript import.

Use cases

  • Keeping project architecture and past decisions available across days of work with Claude Code or Cursor.
  • Sharing one memory store between several coding agents used on the same codebase.
  • Recalling previously fixed bugs and team preferences without copying notes into prompts.
  • Replaying and reviewing what an agent did in earlier sessions.

Quick start

  1. 1

    Run the interactive setup and start the memory server

    npx -y @agentmemory/agentmemory@latest
  2. 2

    Seed sample sessions and test recall

    npx -y @agentmemory/agentmemory@latest demo
  3. 3

    Install the 17 native skills for your agent

    npx skills add rohitg00/agentmemory -y

Things to know

  • Native Windows needs the pinned iii-engine v0.22.1 installed manually; WSL2 or Docker Desktop are easier.
  • Keyless mode disables vector embeddings and uses BM25 for recall unless a local or other embedding provider is configured.
  • Automatic iii-engine installation on macOS/Linux requires curl, sh and tar, which minimal images may lack.
  • It pins iii-engine v0.22.1 and will not work with a different self-installed iii engine version.

Topics

#ai-agents#memory#mcp#coding-assistant#llm#vector-search#claude-code#cursor#agentmemory#agents

FAQ

What is agentmemory?
Persistent memory for AI coding agents that captures sessions automatically and injects relevant context into later sessions across Claude Code, Cursor, Codex and other MCP clients. agentmemory is a memory server for AI coding agents. It silently records what an agent does during a session, compresses those observations into searchable memory, and feeds the relevant context back when the next session starts, so developers do not have to re-explain architecture, past bugs, or preferences each time.
Is agentmemory free to use? What license does it have?
agentmemory is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is agentmemory still maintained?
Actively maintained. The last commit to agentmemory was on Oct 19, 2026.
What can I use agentmemory for?
Keeping project architecture and past decisions available across days of work with Claude Code or Cursor. Sharing one memory store between several coding agents used on the same codebase. Recalling previously fixed bugs and team preferences without copying notes into prompts. Replaying and reviewing what an agent did in earlier sessions.
How do I install agentmemory?
Run the interactive setup and start the memory server: npx -y @agentmemory/agentmemory@latest Seed sample sessions and test recall: npx -y @agentmemory/agentmemory@latest demo Install the 17 native skills for your agent: npx skills add rohitg00/agentmemory -y
Where can I get agentmemory?
The source code and installation instructions are on GitHub: https://github.com/rohitg00/agentmemory

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects