scritty

scritty is a terminal emulator built for the age of AI agents. If you use tools like Claude Code, OpenAI Codex CLI, GitHub Copilot CLI, or Aider, you've probably experienced the frustration of context getting lost between sessions. scritty solves that by sitting between you and your agents. Every time you run an AI CLI inside it, scritty detects which agent is running, captures the entire exchange, tags it with the provider, and indexes it for search. The result is a single, searchable memory bank that lives on your machine. No copy-pasting, no silos.

What makes scritty stand out is how it handles that memory. It's not just a log viewer. The captured data is stored as a vector and keyword index that both you and your agents can query. Agents access it via an MCP server (over stdio or HTTP), so they can search their own and other agents' past conversations. You can search from the terminal with Ctrl+Shift+M, from a browser on the same network, or even from your phone after scanning a QR code. scritty also lets you define standing instructions in a prompt.toml file that gets injected into every agent automatically, so you don't have to repeat yourself. It reads vendor-specific rule files too.

This tool is for anyone who regularly works with multiple AI coding agents and wants to stop losing context. Developers juggling Claude for architecture, Codex for tests, and Copilot for boilerplate will find it especially useful. It also works offline with local models like Ollama, and the browser sync and phone pairing make it easy to monitor long-running agents from anywhere. If you've ever wished your AI tools could remember what they discussed last week, scritty is worth a look.