How it works

Everything runs on your machine. The MCP client and the board talk to the mindpm server; the server and mindpm verify share one SQLite file.

flowchart TB
    subgraph you["You and your agents"]
        direction LR
        client["MCP client<br/>Claude Code, Cursor, ..."]
        board["Kanban board<br/>localhost:3131"]
        verify["mindpm verify<br/>optional, own terminal"]
    end
    server["mindpm server"]
    subgraph disk["On disk"]
        direction LR
        db[("memory.db<br/>SQLite")]
        repo[("your git repo")]
    end

    client -->|MCP tools over stdio| server
    board -->|HTTP + UI token| server
    server -->|read / write| db
    server -.->|git history for the brief| repo
    verify -->|verifier key| db
    verify -.->|checkout of the submitted commit| repo

A typical session:

1. You start a conversation and mention your project
2. The LLM calls start_session → gets full context
3. During the conversation, it creates tasks, logs decisions, adds notes
4. When you're done, it calls end_session → saves what's next
5. Next conversation: instant context, zero re-explanation

What it tracks

• Tasks — handoff status, priority, blockers, sub-tasks
• Specs — objective, why, approach, constraints and acceptance criteria an agent can execute against
• Attempts — every agent run on a task: outcome, root cause, what to avoid next time
• Verification runs — what a verifier executed on a submitted commit (checks, exit codes, test reports) and what it concluded for each acceptance criterion, with evidence
• Verifiers — the registered processes allowed to verify work, identified by a secret key stored only as a hash
• Decisions — what was decided, why, what alternatives were rejected
• Notes — architecture, bugs, ideas, research
• Context — key-value pairs (tech stack, conventions, config)
• Sessions — what was done, what's next