agentmemory

What is agentmemory?

agentmemory is a persistent memory engine for AI coding agents. It runs as a local server next to your agent, hooks into the agent's tool use, and records what happened in each session: file edits, shell commands, test results and errors. Those observations are compressed into durable facts and indexed with a hybrid search that mixes BM25 keyword matching, vector similarity and a knowledge graph. When the next session opens, only the most relevant context is injected within a token budget, so you stop re-explaining your codebase and past decisions. Over the Model Context Protocol (MCP) it serves Gemini CLI, Codex CLI, GitHub Copilot CLI, Cursor, Claude Code and other agents, and one memory store is shared across all of them. The project is Apache-2.0 licensed with no external database to run, so there is no pricing; the only possible cost is the LLM provider you choose for compression, and local embeddings keep that at zero.

Use cases

• Keep codebase context between agent sessions
• Share memory across several coding agents
• Recall past bug fixes and decisions
• Add long-term memory to an MCP client
• Import Claude Code transcripts into memory
• Export agent memories to Obsidian notes

Who it's for

• Software developers
• Claude Code users
• Cursor users
• AI agent builders

Why use agentmemory

• Your agent starts each session already knowing past decisions, so you spend less time restating context.
• Only relevant memories are injected within a token budget, which keeps prompts smaller than replaying full history.
• One local store serves every connected agent, so switching between Claude Code and Cursor does not lose context.
• Data stays on your machine and the engine needs no Postgres, Redis or vector database.
• Capture and recall work without any LLM key, so you can start at no cost.

agentmemory features

• 12 auto-capture hooks across the session lifecycle
• Hybrid recall using BM25, vector search and a knowledge graph
• MCP server exposing 54 tools
• 138 REST endpoints mirroring the MCP tools
• Memory versioning that supersedes near-duplicates
• Lessons with confidence that grows or decays
• Real-time viewer on port 3113
• JSONL session import from Claude Code transcripts
• Peer-to-peer memory sync between nodes over HTTPS
• Obsidian markdown export
• Runs as one process with no external database

agentmemory integrations

• Claude Code
• Cursor
• GitHub Copilot CLI
• Codex CLI
• Gemini CLI
• OpenCode
• Obsidian

How to get started with agentmemory

1. Start the server with npx @agentmemory/agentmemory or a global npm install.
2. Run the connect command to wire it to Claude Code, Cursor or another detected agent.
3. Restart the agent so it picks up the hooks and MCP server.
4. Work as normal while hooks capture prompts, tool calls and results as observations.
5. Open a later session and the most relevant memories are recalled automatically; check them in the viewer or with a smart-search request.

agentmemory limitations

• Needs Node.js 18 or newer
• One-command install targets macOS and Linux; Windows runs via WSL2 or a manual engine step
• Compression, summaries and graph extraction need an LLM provider key

agentmemory pricing

agentmemory is open source, so you can use and self-host it for free.

agentmemory plans

• Open source

  Free

  Apache-2.0, self-hosted

  • Full memory engine
  • Local embeddings option
  • Viewer and REST API

Plans as listed on the official pricing page on 24 Oct 2026. Prices can change and may vary by country or exclude tax. Check the official pricing page.

agentmemory alternatives

The main agentmemory alternatives listed on Debuly are Mem0, Zep, Letta, Supermemory, Cognee and ReMe. The table compares pricing and platforms; order is A–Z.

1. 

   Cognee

   Open-source memory engine for AI agents that combines vector search, knowledge graphs and relational storage, with a hosted cloud.

   Developer platformsAutomation and agentsResearch and search
   FreemiumVisit Cognee (official site)
2. 

   Letta

   Platform, formerly MemGPT, for building stateful AI agents with persistent memory, from a terminal app, SDK or API.

   Developer platformsAutomation and agentsCoding
   FreemiumVisit Letta (official site)
3. 

   Mem0

   Memory layer for AI agents and apps, available as an open-source package or a managed platform that keeps user context across sessions.

   Developer platformsAutomation and agentsCoding
   FreemiumVisit Mem0 (official site)
4. 

   ReMe

   Open-source memory layer for AI agents that stores knowledge as local Markdown files with hybrid search and a linked graph.

   Developer platformsAutomation and agentsCoding
   Open sourceVisit ReMe (official site)
5. 

   Supermemory

   Memory and context API for AI agents with user profiles, a memory graph, retrieval and connectors, billed by usage.

   Developer platformsAutomation and agentsCoding
   FreemiumVisit Supermemory (official site)
6. 

   Zep

   Managed memory and context platform for AI agents, built on the Graphiti knowledge graph, with Python, TypeScript and Go SDKs.

   Developer platformsAutomation and agents
   FreemiumVisit Zep (official site)

agentmemory FAQ

Sources & verification

Facts on this page come from the official sources below, each with the date we read it. Written by Debuly; the vendor did not review this page.

Image credits

• Logo: from agent-memory.dev · fetched 23 Oct 2026 · Vendor trademark, used only to identify the product; removed on request
• Cover: none collected yet, so a generated tile is shown.

Spot an error? Report an issue with this listing. We also remove a logo or image on the owner's request.