claude-mem
AIPersistent memory compression plugin that captures what coding agents do in each session and injects relevant context into future sessions.
claude-mem
Persistent memory compression plugin that captures what coding agents do in each session and injects relevant context into future sessions.
About claude-mem
Claude-Mem is a persistent memory system built for Claude Code that preserves context across sessions. It automatically captures tool usage observations through lifecycle hooks, generates semantic summaries with AI compression, and stores them in a local SQLite database with a Chroma vector index for hybrid semantic and keyword search. At the start of later sessions the relevant context is surfaced automatically, so the agent keeps continuity about a project even after sessions end or reconnect. It also exposes MCP search tools that follow a token-efficient three-layer workflow, and installs with a single npx command or through the Claude Code plugin marketplace. Support extends to other agent hosts such as OpenCode, T3 Code, Antigravity CLI and OpenClaw gateways.
Key Features
Use Cases
Frequently asked questions about claude-mem
What is claude-mem?
Persistent memory compression plugin that captures what coding agents do in each session and injects relevant context into future sessions.
How does claude-mem work?
claude-mem works by combining Automatic Session Capture: Five lifecycle hooks (SessionStart, UserPromptSubmit, PostToolUse, Stop, SessionEnd) record tool usage and observations without manual effort., AI Memory Compression: Generates semantic summaries of what the agent did so past work is stored compactly., Context Injection: Relevant context from previous sessions appears automatically in new sessions., Hybrid Search: A SQLite store plus Chroma vector database enables semantic and keyword search across sessions., Token-Efficient MCP Tools: A three-layer search, timeline and get_observations workflow filters before fetching details, saving about 10x tokens. to help users with Long-Running Projects: Keeping a coding agent aware of decisions and bugfixes made weeks earlier., Session Handoff: Resuming work after a disconnect without re-explaining the codebase., Team Agent Workflows: Preserving institutional knowledge across many agent sessions., Debug History Lookup: Searching what an agent tried previously on a recurring bug., Token Cost Reduction: Pulling only filtered memory details instead of re-reading large context..
What are the main features of claude-mem?
Key features include Automatic Session Capture: Five lifecycle hooks (SessionStart, UserPromptSubmit, PostToolUse, Stop, SessionEnd) record tool usage and observations without manual effort., AI Memory Compression: Generates semantic summaries of what the agent did so past work is stored compactly., Context Injection: Relevant context from previous sessions appears automatically in new sessions., Hybrid Search: A SQLite store plus Chroma vector database enables semantic and keyword search across sessions., Token-Efficient MCP Tools: A three-layer search, timeline and get_observations workflow filters before fetching details, saving about 10x tokens..
Who is claude-mem for?
claude-mem is useful for anyone interested in Long-Running Projects: Keeping a coding agent aware of decisions and bugfixes made weeks earlier., Session Handoff: Resuming work after a disconnect without re-explaining the codebase., Team Agent Workflows: Preserving institutional knowledge across many agent sessions., Debug History Lookup: Searching what an agent tried previously on a recurring bug., Token Cost Reduction: Pulling only filtered memory details instead of re-reading large context..
How much does claude-mem cost?
claude-mem offers a free tier with paid plans for advanced features.
How do I get started with claude-mem?
Visit https://github.com/thedotmack/claude-mem to sign up and explore claude-mem.
Explore more AI Ai Tools tools
Browse all Ai Tools tools →Compare claude-mem: vs Tonefold · vs NOVA CLI · vs Unprompt · vs Velozity