riponcm/projectmem ? reverse-engineered prompt

Build me a Python tool called projectmem that gives AI coding agents a local memory layer inside a repo.

I want it to save what happened while working on code, like issues, failed attempts, fixes, decisions, and notes, then show that back to the agent at the start of the next session so it does not repeat the same mistakes. It should be 100 percent local, with everything stored in a .projectmem folder in the project, no cloud, no account, and no telemetry.

Please include a simple command to set it up in a repo, plus a native MCP server so tools like Claude Code, Cursor, Antigravity, and Codex can read and write memory. I also want the git hook flow so it can warn before a repeated bad approach gets committed, and a small dashboard view for the project memory. If you need to check current MCP or Python docs online, go ahead.