What happened
On 6 October 2025, Zed announced that JetBrains is co-developing the Agent Client Protocol (ACP) — an open standard that lets any AI coding agent work inside any supporting editor, framed as a 'Language Server Protocol for AI agents.' As Zed puts it, 'agents implement one protocol and work everywhere; editors adopt one protocol and support every agent.' JetBrains plans to bring ACP across its IDE lineup (IntelliJ IDEA, PyCharm, WebStorm and more), and in January 2026 the two shipped a shared ACP Agent Registry — a directory of AI coding agents built into JetBrains IDEs (2025.3+) and Zed. ACP supports both local and remote agents.
Why it matters
AI coding agents have been tied to specific editors, forcing an N×M mess of custom integrations. A shared protocol — backed by two major editor vendors — decouples the agent from the IDE the way LSP decoupled language tooling, so the choice of agent and the choice of editor stop being the same decision.
This is the same standardization wave as MCP, one layer up: MCP standardizes how agents reach tools and context; ACP standardizes how agents reach the editor. For teams, the Staff angle is optionality and lock-in — favoring ACP-compatible agents and editors keeps your AI tooling swappable as the market churns, instead of betting the workflow on one vendor's roadmap.
What problem does a protocol like ACP (or LSP) actually solve, and why does standardizing the editor–agent boundary reduce lock-in?
Probes systems thinking about interfaces, N×M integration cost, and platform strategy.