Agent Client Protocol
You can now interact with third-party agents directly within Zed. To make this possible, we created the Agent Client Protocol (ACP), and we've partnered with Google to integrate Gemini CLI as the initial reference implementation.
...
Command-line agents are cool because their simplicity makes them easy to run anywhere—including as a subprocess of another application. Zed was already running Gemini CLI inside our embedded terminal emulator, but we needed a more structured way of communicating than ANSI escape codes. So we defined a minimal set of JSON-RPC endpoints to relay user requests to the agent and render its responses. The result is the Agent Client Protocol, a lean framework that lets any client talk to any agent, as long as they follow the schema.
This Agent Client Protocol is intriguing to me. In some ways it feels like a really natural step forward — I see a significant number of programmers reporting that headless Claude Code gives them better results than in-editor implemented agents like Cursor or Copilot, so text editors getting out of the business of tight coupling agents in them seems like a win. Text editor developer teams don't have to suddenly become agent-coding experts in order to support llm agents. Coding agents can be developed by teams solely focused on their improvement without needing to worry about building a front end. And developers can mix and match their editor of choice with their agent of choice.
It's nice to see they passed the first hurdle, by getting a real coding agent to use the protocol. What remains is for other coding agents to implement the protocol as well. I'm sure the Zed team has thought through the strategy quite a bit, but I'm a little surprised that they think the strategic next step is to get another text editor (Neovim) on board rather than another coding agent. Time will tell whether the strategy succeeds or not.
Assuming the protocol, which I haven't investigated deeply, is effective, I would love for this approach to succeed. My long time love for Emacs, and the unpleasantness of having to learn a new text editor (VSCode/Cursor) just to try an effective coding agent, leads me to want this to succeed. I've tried out Zed and like some of their principles, certainly more than VSCode (which is encumbered by being owned by Microsoft).
This approach also seems to align well with the projected future discussed in the Death of the IDE discussion. Reviewing output code will become important, so being able to use the editor you're most comfortable with, while using it to connect and disconnect from multiple agents working in parallel would be important in that future.