Building an AI agent extension today usually means building the same thing twice — once as an Agent Skill, once as an MCP server, each with its own metadata and discovery path. Vercel just released Agent Plugins 1.0.0, an open standard that makes that redundant.
One manifest file and one directory structure, and your extension works across ChatGPT, Codex, Cursor, GitHub Copilot, Kiro, and VS Code. AWS, Microsoft, OpenAI, GitHub, and Vercel all back it.
What Happened
Agent Plugins 1.0.0 defines a single, predictable directory structure. You write one manifest file and the same package runs on every compatible client, with no per-platform rewrites.
The launch partners read like a who's who of AI: ChatGPT, Codex, Cursor, GitHub Copilot, Kiro, and VS Code. AWS, Microsoft, OpenAI, GitHub, and Vercel all collaborated on the standard, so this is a universal contract rather than a proprietary walled garden.
Why It Matters
For agent developers, this means packaging code once and running it everywhere. The repetitive work of maintaining separate extensions for each client is gone, and AI development moves toward something standardized, portable, and open.
Comments
Join the conversation — sign in to comment.
No comments yet — start the conversation!