Can See

Can See

Let MCP agents see and control terminal apps with screenshots and key input.

Visit Can See

About Can See

Can See is an MCP (Model Context Protocol) server that enables AI agents (such as Claude Code and other MCP-compatible clients) to visually interact with terminal and CLI applications. By launching terminal apps in a virtual terminal and providing capabilities like PNG screenshots, key input, and state comparison, Can See allows agents to debug, test, and operate text-based interfaces based on what 'they see'. The server integrates with CLI tools and the MCP protocol, making it valuable for developers, researchers, and anyone building or testing agent-driven automation for terminal-based apps.

Resources

Product Website

Visit Can See's official website for product details and getting started.

Visit website →