claude-mem: Preserve Agent Context Across Sessions

Overview

Claude-Mem is a persistent memory system for coding agents. It records session activity, creates summaries, and makes relevant observations available in future sessions, helping agents maintain continuity across work that spans multiple conversations.

It is designed primarily for Claude Code and also documents integrations with other agent tools. Use it when an agent needs project history over time and you want searchable, selectively injected context rather than manually repeating prior decisions and findings.

Key Features

• Captures observations from agent sessions through lifecycle hooks or supported integrations.
• Summarizes activity and stores session history in a local SQLite database.
• Retrieves memories through search tools, including a progressive workflow from compact results to full details.
• Combines semantic and keyword search using Chroma and SQLite full-text search.
• Provides a local worker service with a web viewer and search endpoints.
• Supports privacy exclusions using <private> tags and configurable context injection.
• Offers integrations for Claude Code and other agent environments described in the project documentation.

Use Cases

• Developers working across multiple coding sessions can restore relevant implementation history and decisions.
• Teams using coding agents on long-running projects can search prior observations instead of rebuilding context manually.
• Agent users who need to limit prompt overhead can inspect compact search results before loading full memory details.
• Users of supported agent environments can explore persistent memory without building their own session-capture and retrieval system.

Getting Started

For the standard installation, run:

npx claude-mem install

See the README and official documentation for setup options and configuration.

Alternatives

• ruflo: Ruflo bundles persistent memory with multi-agent coordination and workflow automation, while claude-mem focuses on capturing and retrieving Claude Code session history.
• Memori: Memori provides structured memory SDKs for general LLM applications, while claude-mem targets coding-agent activity and project history in Claude Code.

Considerations

• The README lists Node.js 20 or higher as a requirement. Bun and uv are used by the system and are auto-installed if missing.
• Installation may prompt users to sign in and choose a memory provider. The README says this can be skipped with an explicit provider, an opt-out setting, or a non-interactive install.
• Memory capture and context injection depend on the agent integration and its configuration. Review privacy exclusions and settings before storing sensitive project context.
• The project is actively maintained according to the supplied repository metadata, but its large feature set and multiple integrations may require checking the documentation for the specific agent environment you use.