MenteDB MCP
MenteDB keeps agent memory in a Rust database: an MCP server for Claude Code and Cursor in cloud or self-hosted mode.
- 126 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- Conversations and extracted entities are stored in the database and may go to the cloud or an external LLM provider.
- Cloud mode requires an access key.
Install
- Works in
Manual install
npx mentedb-mcp@latest setup claude-codeREADME command, and `claude-code` can be replaced with cursor, copilot or claude.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Conversations and extracted entities are stored in the database and may go to the cloud or an external LLM provider.
- Cloud mode requires an access key.
This is third-party code. Review the repository files before installing.
What it does
MenteDB gives an AI agent long-term memory. It is a Rust engine with a write-ahead log, an HNSW index and a knowledge graph, and it extracts people, places and events from conversations and links them without dumping every phrase. The assistant calls `process_turn` every turn and curated context lands in the prompt. Connecting a client is done by `npx mentedb-mcp@latest setup claude-code`, and Cursor, Copilot and Claude work too. There is a cloud mode with the `MENTEDB_API_KEY` key and a self-hosted mode: `cargo install mentedb-server` or Docker, REST on port 6677. SDKs for Python, Node and Rust exist for custom agents.
Who it is for. Agent builders and people who want Claude Code or Cursor to remember people, projects and past decisions across sessions.
Good fit when
- An assistant has to remember preferences and project facts across sessions.
- Memory is needed on your own server, with no cloud.
- A custom Python, Node or Rust agent needs a ready memory layer.
- Memory queries should return curated context without the whole archive.
Not a fit when
- A notes file or `CLAUDE.md` is enough.
- Data cannot go to a cloud service and a server cannot be deployed.
- You need an ordinary relational database.
Example request
Remember that this project uses pnpm and that we leave the legacy folder alone, and remind me in the next session.MCP
- Transport
- stdio, http
- Authentication
- key is optional
Environment variables
MENTEDB_API_KEYsecret- MenteDB cloud mode key
Limitations
Cloud mode needs an account on app.mentedb.com and the `MENTEDB_API_KEY` key, and conversations go to their server. Self-hosted mode needs Rust or Docker, and entity extraction uses an external LLM provider key, such as OpenAI, and sends texts to it. Memory keeps facts about people and events from conversations, so restrict access to the data directory. The project is young and the data format and API may change.
How to disable. Remove the MenteDB entry from the MCP client settings, stop the server or container and delete the data directory. Revoke the cloud key in app.mentedb.com.
FAQ
Which clients connect?
Claude Code, Cursor, Copilot and Claude through `npx mentedb-mcp@latest setup` with the client name.
How does cloud differ from self-hosted?
Cloud is managed by the service and needs a key. Self-hosted runs on your machine through `cargo install mentedb-server` or Docker, with nightly maintenance built in.
How do I run my own server?
With `mentedb-server --data-dir ./data`, and REST will listen on port 6677.
What does process_turn do?
It embeds the user message, recalls what is relevant and returns context for the prompt.
Is there an SDK?
Yes, for Python (`pip install mentedb`), Node and Rust, plus memory providers for LangChain and CrewAI.
Discussion
No questions or reviews yet.
Related
MCP serversDurable memory for agents: an MCP daemon with a hash-chained thought log and a shared versioned skill registry.
Agent memory in Rust: forgets the unimportant, merges similar items, flags contradictions, runs locally, over HTTP and via MCP.
MemMesh, MCP-сервер памяти в одном бинарнике на Rust, хранит заметки агента на машине и ищет по смыслу без внешней базы.
A long-term memory MCP server for agents: one Rust binary that runs offline and never calls a language model
Local memory for AI agents: Pensyve keeps facts, conversations and action outcomes in SQLite and returns them in later sessions.
Self-hosted long-term memory for agents: semantic search, a knowledge graph and autonomous consolidation over MCP and a REST API
README in short
The description shows two run modes, installation through Docker, Rust and the Python SDK, then the memory curation principle and comparison results. Ready agent examples exist for several languages.