Engineering brief
AI coding agents just got their LSP—and it’s not from Microsoft
At a glance
- Relevance
- Practical value
- Warnings
- None
Zed’s Agent Client Protocol (ACP), now supported by 40+ clients and agents including JetBrains and Cursor, could decouple coding agent choice from editor lock-in. However, remote transport is still missing and enterprise governance remains unaddressed.
Standardizing agent-editor communication could cut integration overhead and let teams choose best-of-breed agents independently from editors.
Summary
Zed’s live ACP demo showed a minimal coding agent easily becoming compatible via its JSON-RPC based protocol, now supported by 40+ clients and agents including JetBrains, Obsidian, Cursor, and OpenClaw. The session featured streaming model output, tool calls, and proxied file system access—all decoupled from the editor.
The immediate implication for engineering leaders is a potential de-facto standard for agent-editor interaction. If ACP follows the trajectory of LSP, teams gain the ability to swap coding agents without changing editors, reducing vendor lock-in and integration cost. It separates the agent ecosystem from the editor ecosystem.
Yet there are tradeoffs. ACP adds a new layer of protocol complexity alongside MCP and LSP. Remote transport is still under development, and enterprise-grade features like security governance are unaddressed. The demo’s simplicity is promising, but production readiness remains unproven.
What to watch: adoption by VS Code and other major editors will be pivotal. ACP’s built-in ability to proxy unsaved editor buffers and terminals could introduce subtle consistency and security risks that engineering orgs should evaluate before mandating its use.
Watch the video
This video is blocked due to your privacy settings. To watch this video, please accept YouTube marketing cookies.
Related breakdowns
MCP apps fail without data-model separation — Indeed's practical lesson
Indeed's AI team reveals why most MCP app implementations create a black-box problem. The fix: separate data processing from UI rendering to keep the model…
Why AI-generated code needs mathematical proof, not just tests
AI coding agents generate PRs faster than humans can review. Formal verification with Lean mathematically proves correctness for all inputs—something tests…
Why Frontier LLMs Still Can't Write Fast Multi-GPU Kernels
LLMs solve only a third of multi-GPU kernel tasks despite excelling on single-GPU benchmarks. The bottleneck has shifted to communication, and reasoning…
Get TL;DW
Too Long; Didn't Watch.
A concise breakdowns of the AI and devtools videos that actually matter for engineering leaders.
Free. Weekly. No hype.
Video and thumbnails remain the property of their respective creators. tldw.news provides editorial analysis, commentary, and discovery links to original content.