mindpm — Claude Code MCP server: project memory and task control
mindpm — an MCP server that stores tasks, decisions and project context in SQLite. Every new conversation picks up where the last one left off.
Apache-2.0 ★ 9 updated 2026-10-05
claude mcp add --env MINDPM_DB_PATH=<MINDPM_DB_PATH> --env MINDPM_PORT=<MINDPM_PORT> --env MINDPM_HOST=<MINDPM_HOST> --transport stdio mindpm -- npx -y mindpmWhat mindpm does
mindpm writes project state to a local SQLite database: tasks with priorities and blockers, accepted decisions and rejected alternatives, architecture notes, session summaries and agent attempts. The LLM calls start_session to load context, adds new records as you work, and saves the next-session plan via end_session. A Kanban board at localhost:3131 displays tasks and lets you accept agent work.
The server can hand specced tasks to autonomous agents: a claim mechanism with leases prevents parallel agents from working on the same task, and you accept results manually on the board. For unattended agents, an optional verifier reruns checks before counting a commit as done. All data stays on your machine in a single ~/.mindpm/memory.db file.
No user accounts, no write protection — an agent with shell access can bypass the board and write to the database directly. Review the security model in the docs before running agents unattended.
What it helps with
- preserve project context between sessions in Claude Code
- hand a specced task to an agent without chat history
- stop two agents from working on the same task at once
- run a Kanban board of tasks and accept agent work
- rerun checks before counting an unattended agent's commit as done
Example prompts for the agent
- Show high-priority tasks from mindpm
- Save the decision to use Redis for session cache in mindpm
- Write the session summary and next-session plan to mindpm
How to install mindpm
Platform CLI
claude mcp add --env MINDPM_DB_PATH=<MINDPM_DB_PATH> --env MINDPM_PORT=<MINDPM_PORT> --env MINDPM_HOST=<MINDPM_HOST> --transport stdio mindpm -- npx -y mindpmReplace <…> with your own values
Platform CLI
codex mcp add mindpm --env MINDPM_DB_PATH=<MINDPM_DB_PATH> --env MINDPM_PORT=<MINDPM_PORT> --env MINDPM_HOST=<MINDPM_HOST> -- npx -y mindpmReplace <…> with your own values
Config file
{
"mcpServers": {
"mindpm": {
"command": "npx",
"args": [
"-y",
"mindpm"
],
"env": {
"MINDPM_DB_PATH": "<MINDPM_DB_PATH>",
"MINDPM_PORT": "<MINDPM_PORT>",
"MINDPM_HOST": "<MINDPM_HOST>"
}
}
}
}Replace <…> with your own values
Config file
{
"servers": {
"mindpm": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"mindpm"
],
"env": {
"MINDPM_DB_PATH": "<MINDPM_DB_PATH>",
"MINDPM_PORT": "<MINDPM_PORT>",
"MINDPM_HOST": "<MINDPM_HOST>"
}
}
}
}Replace <…> with your own values
Platform CLI
gemini mcp add -e MINDPM_DB_PATH=<MINDPM_DB_PATH> -e MINDPM_PORT=<MINDPM_PORT> -e MINDPM_HOST=<MINDPM_HOST> mindpm npx -y mindpmReplace <…> with your own values
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about mindpm
How do I install mindpm in Claude Code?
Run claude mcp add mindpm -- npx -y mindpm. To set the database path and board port, add --env MINDPM_DB_PATH=<path> and --env MINDPM_PORT=<port> after add.
What does mindpm do?
It stores tasks, decisions, architecture notes and session summaries in SQLite. The LLM reads context via MCP tools at session start and writes changes at the end, without relying on chat history.
Where does mindpm store data?
In a SQLite file on your machine, ~/.mindpm/memory.db by default. Set the path with the MINDPM_DB_PATH environment variable.