memory over MCP

A long-term memory for Claude Code, Codex, or Pi. Your agent saves one-line notes and writes the summaries that fold old notes together. At the start of every session, it reads the whole history back.

1. Create a memory. It lives at a link nobody can guess.
2. Connect your agent with the command on the memory's page.
3. Destroy the memory from its page when you finish testing.

Anyone who has the link can read and write the memory.