95,920 stars · Apache-2.0 (read from /blob/main/LICENSE; stock unmodified text, BUT the appendix copyright line is left as the template "Copyright [yyyy] [name of copyright owner]" and no holder is asserted anywhere in the file) · v13.29.0 (2026-10-04; the GitHub page shows "03 Oct" with no year and the year was settled by registry.npmjs.org, which timestamps 13.29.0 at 2026-10-04) · Track this in Scout
It gives a coding assistant a memory that survives closing the window, stored in a local SQLite file.
▶Repo detailsthe review · specs · pros & cons · install
What it is
claude-mem is a plug-in for Claude Code that stores session context in a local SQLite database (a database that is one ordinary file, with no server to run) and restores it on the next session. It is thirteen months old and already at version 13.
What it is good for. Anyone who works with a coding assistant across several days on the same codebase and is tired of explaining the same decisions every morning. It is less useful to someone who opens an assistant for one question and closes it.
- One command,
npx claude-mem install, and it is also listed in the Claude Code plug-in marketplace. - Apache-2.0, read from the file at
/blob/main/LICENSE, and the text is the plain unmodified licence with no added conditions. - Code was pushed on 4 October 2026, the morning this edition was written, and version 13.29.0 reached the npm registry on the same day.
- This is the fact this edition is named after. The Apache licence file is the stock text, and the copyright line was never filled in. It still reads
Copyright [yyyy] [name of copyright owner]. On a project with 95,920 stars, nobody is named as the person or company granting the licence. The licence still works, but anyone doing legal checks before using it commercially has nothing to point at. - Version 13 in thirteen months, with 13.24.1 and 13.29.0 landing inside one month. Pinning a version is advisable, and expect breaking changes.
- It installs a second program runtime called Bun on its own, writes a local database, and offers an optional cloud sync. 52 open pull requests against 32 open issues means more changes are waiting than problems are.
mem0ai/mem0A memory layer for AI applications in general rather than for one assistant, and a library you write code against rather than a plug-in you install.
Track this in Scout- doobidoo/mcp-memory-service
The same idea delivered as an MCP server, so it works with any assistant that speaks that protocol, and correspondingly more to set up.
Track this in Scout
npx claude-mem install
