When you work with a coding agent like Claude Code, the context disappears every time you close a session. The next day you explain the same premises all over again, or you can no longer recall why a particular design decision was made. Claude-Mem sets out to solve exactly that, through automatic recording of your work and full-text search over it. It is developed by thedotmack, released as open source under the Apache-2.0 license, and installs with a single command: npx claude-mem install. A background observer records the decisions, bug fixes, and architectural choices made during a session as structured “observations”, making them searchable and referenceable in later sessions.
Key Features
- Automatic recording across sessions: It hooks into the SessionStart / UserPromptSubmit / PostToolUse / Summary / SessionEnd lifecycle events and accumulates tool-execution history and summaries automatically. You never have to perform a “take a note” action yourself
- Fast full-text search with SQLite + FTS5: Records are stored in a local SQLite database and indexed with FTS5 for full-text search. Optionally, combining it with Chroma enables meaning-based vector search
- Natural-language search: Through MCP tools and the bundled skills, you can pull up past context with plain sentences such as “what did we decide about authentication last week”. It also shows citations back to the original observations
- Automatic CLAUDE.md generation: It generates and updates a
CLAUDE.mdfile summarizing the project’s activity timeline, so it can be handed over as context when the next session starts - Privacy controls: Anything wrapped in
<private>tags is excluded from recording. Data stays local by default - Multiple client support: Beyond Claude Code, it can be used from Cursor, Windsurf, OpenCode, Codex CLI, Gemini CLI, OpenClaw, VS Code, and other MCP-capable agents
- Web viewer: A bundled management UI lets you watch the stream of recorded memories in the browser in real time
Pricing
| Plan | Price | Highlights |
|---|---|---|
| claude-mem (open source) | $0 | All features available locally. Apache-2.0 license |
| CMEM Cloud | $20 / month | Cloud sync, plus a private MCP endpoint accessible from anywhere |
| Team | $333 / seat / month (3–50 seats) | Team usage with FDE onboarding support |
Pricing is current as of August 2026. Check the official site for the latest details.
Pros & Cons
✅ Pros
- The core is open source and free. You can keep everything local without touching the cloud
- Recording is fully automatic, so there is no burden of “leaving notes” while you work
- Because it is SQLite-based, the recorded data lives on your machine and you can inspect and back it up yourself
- It is not exclusive to Claude Code — Cursor, Windsurf, and various CLI agents can share the same memory
- The
<private>tag gives you a way to keep sensitive material out of the record
⚠️ Cons
- It assumes a runtime environment with Node.js 20 or later, Bun, and SQLite 3, so it is not a tool non-engineers can pick up casually
- Since everything is recorded automatically, you need operational discipline to keep the store from bloating and to control noise
- Meaning-based search requires the extra setup of Chroma
- Releases are frequent, which means some ongoing cost in keeping up with breaking changes and version differences
- Cloud sync and team sharing sit behind paid plans, and the Team tier is priced out of reach for individuals
Comparison with Similar Services
| Criteria | Claude-Mem | Mem0 | Basic Memory | CLAUDE.md (hand-written) |
|---|---|---|---|---|
| Main use | Automatic work-history recording for coding agents | Memory layer embedded in applications | Markdown-based knowledge note integration | Manually written project premises |
| Recording method | Automatic (via hooks) | Depends on the app implementation | Semi-automatic (saved as notes) | Fully manual |
| Storage | Local SQLite (plus optional cloud) | Local / managed cloud | Local Markdown files | Text inside the repository |
| Search | Full-text (FTS5) plus optional vector search | Vector search | Full-text and link traversal | No search |
| Cost | Free (cloud is paid) | Free tier / paid plans | Free (open source) | Free |
Who Is It For
- Developers who use Claude Code daily and want to stop re-explaining the same premises every session
- People who want to keep the reasoning behind “why we implemented it this way” in a form they can search later
- Anyone on a long-running project where recalling work from several weeks ago comes up frequently
- Users who run Cursor, Windsurf, and various CLI agents side by side and want shared context across environments
- Teams and individuals who prefer to keep code and work history local rather than sending it to the cloud
Summary
Claude-Mem is an open-source plugin that tackles head-on the structural weakness of coding agents — that memory vanishes when a session ends — using automatic recording and full-text search. Installation takes one command, the core is free, and the data stays in a local SQLite database. On the other hand, it presumes a Node.js and Bun runtime, and the effort of curbing record bloat and keeping up with releases falls on you. A practical path is to start with the open-source version locally, and consider CMEM Cloud once you need to share memory across multiple machines or a team.