Zed Editor Integrates Gemini via Open Agent Protocol

Opening: A new chapter in AI-assisted coding

If you’ve ever watched a coding session feel a little like steering a ship through fog, you’ll understand why the latest collaboration between Zed and Google feels noteworthy. Zed, a high-performance, Rust-based code editor, has wired in Google's Gemini CLI directly into its environment. This isn’t just about adding a shiny tech stack; it’s about rethinking how developers talk to AI while they write code. Gemini isn’t a separate tool you launch and forget about — it’s becoming a native part of the editor itself, guided by a fresh, open protocol for AI agents.

How the new open protocol works

At the heart of the integration is the Agent Client Protocol (ACP), an open-source framework built by the Zed team. Think of it as the editor’s new universal bus for AI agents, modeled after the Language Server Protocol (LSP) that decouples language specifics from the editor. The ACP aims to standardize how agents talk to editors, so you can swap in different AI assistants without reconfiguring your entire toolchain or chasing a vendor lock.

Gemini CLI is the initial reference implementation for ACP, showing what a model-powered workflow can look like inside a fast editor. The Gemini integration demonstrates a couple of core ideas:

• AI is inside, not on the side: You don’t run Gemini as a separate window; you invoke it from within Zed and have it act directly on your code.
• Real-time transparency: The editor’s “Follow Agent” feature lets you watch the AI’s read-and-write actions across multiple files as it tackles a task, so you can see the AI’s execution path before you approve changes.
• Contextual help, on demand: Select a hint of confusing code or a cryptic error, press a hotkey, and Gemini offers a precise, in-context explanation rather than a generic reply.
• In-editor collaboration: A natural-language chat, available from Zed’s integrated terminal, lets you ask broader questions and get help without leaving the editor.

What you can do in practice

The Gemini-enabled workflow inside Zed is designed to feel familiar if you’re used to command-line power wrapped in a polished IDE.

• Generate or refactor in place: Write a descriptive comment and use a hotkey — Gemini fills in boilerplate or implements complex logic right where you’re typing.
• Explain confusing blocks: Highlight code, and Gemini explains it in plain terms, helping you understand breadcrumbs you otherwise might miss.
• Chat with AI, in-terminal: The natural-language interface is accessible from Zed’s terminal, so you can pivot to a broader question without leaving your workspace.
• Review, don’t auto-accept: The integration surfaces changes in a review-style panel, letting you accept, reject, or tweak each suggestion before it lands in your repo.

The open ecosystem, with you in the driver’s seat

ACP is designed to prevent vendor lock-in and to enable a diverse ecosystem of AI assistants. It’s intentionally neutral: you could switch between different agents based on language, project, or personal preference. Gemini is the first reference implementation, but the hope is that other providers will build compatible agents that plug into Zed (and other editors that adopt ACP).

Google and Zed aren’t just testing a feature for a single product; they’re signaling a broader thesis: AI-assisted development shouldn’t force you to live inside a single vendor’s universe. By making ACP open and the Gemini integration readily accessible, both teams are nudging the industry toward a marketplace of AI helpers that work inside the editor you already trust.

Why this matters for the future of coding

This isn’t just a feature bump. It’s a blueprint for how developers and AI can collaborate more transparently and efficiently. With ACP, you get the reassurance that the AI’s path through your code is visible, reversible, and controllable — a practical counterbalance to the speed of machine-generated suggestions. And because Gemini is running inside Zed, you’re not jumping between contexts or fighting misaligned toolchains; you’re coordinating a single, coherent workflow.

The collaboration is also a strategic move for both players. For Google, embedding Gemini in a popular editor helps meet developers where they are, in the environments they already use. For Zed, this underscores its position as a fast, extensible, and neutral platform for AI-assisted programming, capable of hosting third-party agents rather than charging you to stay inside a single ecosystem.

Looking ahead

If the ACP approach catches on, we could see a future where editors become hubs for multiple AI agents, each tailored to different languages, domains, or preferences. Gemini’s role here is as a leading example — a model that demonstrates how a thoughtful, open protocol can lower barriers to experimentation while keeping you in control of every change.

For more background on ACP and Gemini’s reference role, see the sources linked below.

Sources and further reading

• https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGssCrIhXYt-RlMYeT6NXrQRK8siIJXghGIdTAY6UOMKHHitoV-9u3pUOjMz-5irYhNFF708wpHniurIQi7a4jEOKmlilzfDYZZvKbx3kOgVcczcBlr4iHriqrF9UAWcpMrZ_W3c4x7CPsYblGjT9aA2jc_cHSYn7teDos-uYGdOpvqx0r5gg==
• https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHZd98_h4oQaYDElO_IiBSHQ1YFd_0Kd64SJCX6MX7Lexfd20C-gn9bPgtXpd89iVk_HyPizGLWgiH6gt3OIP2KRDVHuJq_X1eTkgt5Hf2yLu0YSfXeWATmyiKIXXLErrrXf3ttIzKKdv3mcXv1StTim6_QiWNHkuepQ4x6T18HeQ9DoK-ORFKNYwvmkmQSWCX1
• https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHESOZNlAK9uE_6tZEv6AgzbFT4poF2qmj8U3k08bqMp3M9Dg9UnF_eZPJ9rdde7MQwmHm-wlMIUzt75NHyd5fFbd6c2t9MZ2_beIxPIMEElGiPl2zInby1cwhHV192J57HB0ow23bhLtfYMGYYCVmtSsxye6kLrGd6rAgtbdRampyOLMzwHXf7B1IpXg==
• https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFfEudk4ezc8IQBBjF7dH8q8G7_WKT81fVR2u9ckYg-e7s2QmE7NBUg39FPznLAcaYKvCLPSf23RLT5EeLeGhquZUzIFZTRVni70Ofa4xpUaCOG2n_XghfaEFTwckj2CucblvWEGdrzKX62wav25CigcLSGE8nxewY=
• https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEU0hf_x27NUeijykUFV8EdGOGl2TrQAJ67fs6fN5xSLmXHhd-Yw2He5Dc2hlAOWWwRtZks50mo37DEkii-8wnnBlyGSXaAIh1AWRq7pdbt_MkFJ5uioRcvsQzFlfJuooVhmfAP3Avx8vDtf11uDXYiolBucpdYn5_jdZ9MT1lFAfy3IZebZzFuPvNGfps7s5pPGF_74bnpUeqYSF6R0GNr4NpQqlgyK3yD59Pxeegg