Cursor, OpenAI, Microsoft, AWS and Vercel agree a shared plugin format — AI coding tools stop being separate silos

Agent Plugins 1.0, published 6 August 2026 by Vercel with AWS, Anysphere (Cursor), Microsoft and OpenAI, is a vendor-neutral package format for AI agent skills and MCP servers that already loads on day one in ChatGPT, Codex, Cursor, GitHub Copilot, Kiro and VS Code — a direct answer to the search for a standard AI coding tool format.

Five of the biggest names in AI coding tooling — AWS, Anysphere (Cursor’s maker), Microsoft, OpenAI and Vercel, who initiated the proposal — published Agent Plugins 1.0 on 6 August 2026. It’s a small, deliberately narrow specification: a folder with a plugin.json manifest that packages an MCP server and a set of agent skills into one installable unit any compatible client can load. Six tools support it immediately — ChatGPT, Codex, Cursor, GitHub Copilot, Kiro and VS Code — with GitHub shipping support the same week.

The problem it solves is one every team that’s tried more than one AI coding tool has hit directly: every agent expects its own folder layout, its own manifest format, its own setup steps. A custom internal tool, an MCP server for your API, a house-style skill for how your team writes tests — under the old world, each of those had to be rebuilt separately for Cursor, then again for Copilot, then again for whatever the team switches to next. Agent Plugins means build once, and the plugin follows you across tools rather than locking you into whichever one you built for first.

That’s the detail worth sitting with if you’ve been avoiding investing in AI tooling because you didn’t want to bet on the wrong vendor. The standard deliberately stays out of installation, marketplaces, permissions and sandboxing — it leaves each client to handle security and distribution its own way — so it’s a packaging format, not a security guarantee. But it does mean the switching cost between AI coding tools just dropped, and the tooling investment you make today has a real chance of surviving your next platform decision.

So what

If you’ve held off building internal AI-agent tooling — a skill for your codebase conventions, an MCP connector into your own systems — because you weren’t sure which coding assistant your team would standardise on, that’s less of a blocker now. We build AI-assisted development workflows around the tools that actually fit a project, not a single vendor’s ecosystem — see our AI-assisted development approach, or get in touch if you want help scoping what’s worth building versus what a plugin can now solve off the shelf.

← All research

More research.

GitHub Copilot adds Supabase, Webflow, Wix and five more connectors in public preview

GitHub Copilot now connects to eight more services, including Supabase, Webflow, Wix and Mixpanel, through verified MCP connectors in public preview. They join more than 25 existing ones.

Read ›

Gemini 4 Neon is generally available in GitHub Copilot, aimed at lower-cost agentic coding

Gemini 4 Neon is now generally available in GitHub Copilot for agentic coding. Google says its quality is close to Gemini 4 Carbon at a lower token cost per task.

Read ›

Claude Code 2.1.328 brings auto mode to Bedrock, Vertex and Foundry, and narrows /code-review to bugs

Claude Code 2.1.328 extends auto mode to Claude Opus 4.6 and Sonnet 4.6 on Amazon Bedrock, Google Cloud and Microsoft Foundry. /code-review now reports only bugs and CLAUDE.md violations.

Read ›

GitHub Copilot Evals: repeatable tests for AI agents, now in public preview

GitHub Copilot Evals lets teams write repeatable tests for AI coding skills, custom agents and MCP tools. It checks both the result and the path the agent took, and is in public preview.

Read ›

GitHub Copilot can now drive your browser: browser use enters public preview

GitHub Copilot browser use lets an AI coding agent control Chrome to debug web apps and run browser tests. It is in public preview across the CLI, the Copilot app, VS Code and JetBrains.

Read ›

Claude Code 2.1.327 refuses some symbolic links and lists broken setup in /status

Claude Code 2.1.327 tightens file and package permissions again and adds a /status list of broken agent, skill and hook settings. A short read for teams using AI coding tools on a live product.

Read ›

Start a project

Building with AI? Let's talk.

Tell us what you are building. We'll give you an honest view of the right approach.

Start a projectSee our work