What it does
CP Memory keeps memory for AI coding agents in a local SQLite database, by default at ~/.cp-memory/memory.db. It stores facts, preferences, ongoing work, episodes, decisions and conversation checkpoints, and restores relevant context through MCP tools. Project-scoped memories are prioritised for the current project. Memories can be reviewed, corrected, marked wrong or stale, or scoped, instead of being silently overwritten, and the correction history is kept. A weekly maintenance task runs health checks and low-risk expiry cleanup.
A user would reach for it to have an agent keep project rules, such as a release process, across sessions, and to fix a stored rule without losing its history. The stdio MCP server is portable and can be used from several MCP clients, including Codex, Claude Code, Cursor, VS Code and Gemini CLI. The Codex plugin adds Skills and lifecycle hooks that restore context on startup and on eligible prompts.
The portable server needs Python 3.10 or later and uv. On Windows, the recommended route is installation from the GitHub Marketplace. On macOS and Linux, a source installer creates a private Python runtime for the plugin and installs the MCP dependency. The publisher reports clean-cache MCP handshake checks and unit tests passing on Windows, macOS and Linux in CI.
Limitations and requirements
Requires Python 3.10 or later and uv for the portable MCP server; the macOS/Linux source installer requires Python 3 on the PATH. Marketplace installation on macOS/Linux does not run install.sh and does not create the private runtime, so it is not an equivalently verified path. Manual smoke testing of real Codex desktop hook injection on macOS/Linux is still pending. Review reminders are injected into assistant context and are not a visible review panel; the tool does not auto-delete memory or auto-resolve conflicts. Do not commit a real memory.db, logs, private summaries or environment files.
Safety
Our team has not reviewed this resource. Only automated rules checked it; resources with high-risk signals are not published. These are signals, not a guarantee that installing it is safe. Try it in a sandbox project first.
- Runs code or changes an assistant's behaviour
Sources
- Official MCP Registry first seen 13 Nov 2026 · last seen 13 Nov 2026 · its page on the source
CP Memory
Local-first, governable long-term memory for AI coding agents
v1.8.1As shown on registry.modelcontextprotocol.io when we last read it