MemoMind MCP
MemoMind MCP gives Claude Code project memory, with stored facts, events and conclusions visible in a local dashboard.
- 602 ★ on the repository
- Last commit
Extended access
Why this level
- The agent reads and writes long-term memory; a cloud LLM receives conversation facts for processing.
Install
- Works in
Manual install
claude mcp add --scope user --transport stdio memomind \
-- D:\memomind-env\Scripts\python.exe D:\memomind-env\mcp_stdio.pyMCP registration from the README’s first, Windows setup path. Prepare the server and dependencies first, then use your own paths.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The agent reads and writes long-term memory; a cloud LLM receives conversation facts for processing.
This is third-party code. Review the repository files before installing.
What it does
MemoMind MCP connects Claude Code to a Hindsight-based long-term memory server. The agent stores facts and retrieves history relevant to its current task. Search uses keywords, meaning, relationships and time, so each session can request context without loading every conversation. Separate memory banks hold history for different projects or users. Through the local dashboard, users browse stored records, search them and export memory as JSON. They can inspect the bank’s contents separately from the agent’s answer.
Who it is for. Teams using Claude Code on long projects where fresh sessions repeatedly need earlier decisions and preferences.
Good fit when
- After a break, the agent asks again about past project decisions.
- A new session makes you repeat preferences you already explained.
- Several projects need separate memory banks.
- When switching models, you want to keep project context in your own bank.
Not a fit when
- You cannot maintain a separate PostgreSQL database.
- History must not reach the selected model provider.
- A short static rules file is sufficient for the task.
Example request
Recall this project’s authentication decisions from MemoMind and distinguish stored facts from model conclusions.MCP
- Transport
- stdio
- Authentication
- not required
Limitations
Prepare Python 3.11+, PostgreSQL 17 with pgvector and the MemoMind server first; the connection command does not install them. Fact extraction uses the selected LLM, so local database storage does not imply local processing of all text. Each MCP process loads its own embedding model and consumes additional memory, especially across concurrent sessions. Conversation import and document scanning can add personal information; enable only required sources and inspect the stored records.
How to disable. Remove MemoMind from Claude Code, stop its server and disable configured startup tasks. Memory databases and backups remain separate.
FAQ
Where can I inspect stored memories?
Open `http://127.0.0.1:9999` while the dashboard is running to browse and search records.
How do I check the server before connecting?
Open `http://localhost:19999/health`. The values `status: healthy` and `database: connected` confirm that the server is running and connected to its database.
Can I export memories?
Yes. The dashboard exports JSON, and the project includes backup tools. Exports contain stored information, so remove unwanted personal data before sharing an archive.
What does Windows setup require?
Native setup needs Visual Studio Build Tools to compile pgvector. A separate WSL2 installation path is also documented.
Can I add older conversations?
Yes. The project supports ChatGPT and Gemini history imports. Inspect the export first and choose what may be retained.
Discussion
No questions or reviews yet.
Related
MCP serversLocal memory for agents: stores conversations and project files verbatim and retrieves them with semantic search over MCP
An MCP server that keeps project decisions in Redis or Valkey and brings them back in new sessions and after compaction.
Long-term memory for Claude Code, Cursor and Codex: the agent saves decisions and finds them again in later sessions
An MCP server that preserves Claude Code context between sessions: decisions, checkpoints and a file cache kept in local SQLite.
A local workbench for Claude Code, Codex, Pi and OpenCode: memory changes wait for your review, then an audit follows each turn
Long-term memory for agents: a remote MCP server, plugins for Claude Code, Cursor and Codex, and an API for your own apps
README in short
Storage distinguishes facts, events, observations and topic-level conclusions. A bank mission gives context for consolidating new records and can describe a particular project or user.