ai-memory

Why we included this project

Switching between coding agents usually means watching the new one forget what the last session already figured out. ai-memory gives those agents a long-term memory for a working directory, storing the architecture you settled on, the approaches that failed, and the open questions so a fresh session in Claude Code or Codex starts with real context instead of a blank slate. It runs as a single Rust binary and plugs into each agent through MCP servers and lifecycle hooks, and it can pass state between different vendors, so you can quit one tool mid-task and resume in another without re-explaining the work. The web UI and managed workstream routing make that memory easy to inspect and adjust, which matters if you juggle several agent harnesses on the same codebase. It is shared infrastructure for keeping agents coherent across sessions, not another single-agent wrapper.