A developer has introduced Terminai—a tool that allows for the seamless integration of CLI agents, such as Claude Code or Codex, into the current terminal session using a transparent PTY overlay and the MCP protocol.
What Happened
The Terminai project has been presented, operating as a transparent shell for the terminal. The tool uses a local MCP (Model Context Protocol) server to pass context—screen content, the current working directory, and command history—directly to the AI agent. To ensure security, all commands proposed by the agent require mandatory user confirmation before execution in the shell.
Context
Unlike specialized terminals like Warp or isolated chat interfaces, Terminai uses a "transparent augmentation" architecture. The project is based on the mprocs terminal emulation mechanisms, allowing AI agent capabilities to be embedded into existing working environments without the need to switch to a different terminal emulator.
Why It Matters for the Industry
The project demonstrates an important shift in UX for development tools: the transition from a separate chat model to an "invisible AI" model integrated into the workflow. The use of MCP (Model Context Protocol) sets a standard for passing context between local tools and LLM agents, promoting standardization of interaction between the shell environment and AI.
Why It Matters for Users
Developers can use their existing AI model subscriptions (e.g., Claude or Codex) directly within their favorite terminal. This eliminates the need for manual context copying and reduces the cognitive load of switching between the terminal and an AI interface, while maintaining full control over commands through an approval system.
What Is Not Yet Known / Limitations
There is a fundamental risk of data leakage through the transmitted context (screen, history), which may not be fully mitigated by the command confirmation system.
Sources
Author
Look at AI, Editorial Team