Agent Plugins

An open standard for packaging agent instructions and tool connections into one folder that works across different AI clients.

A plugin is a folder holding a plugin.json file, a skills folder for reusable instruction sets, and an mcp.json file describing connections to live tools and data
Client-specific settings live in separate folders named after the vendor, so one plugin can carry extras for individual clients without breaking the others
Codex, ChatGPT, Cursor, GitHub Copilot, Kiro, and VS Code support the format at launch
The 1.0.0 specification is openly licensed and steered by a committee with people from Amazon, Cursor, Microsoft, OpenAI, and Vercel
The spec is published at agent-plugins.org

Building a reusable set of instructions and tool connections once instead of rebuilding it for each AI client your team uses.

Your team runs more than one agent client and keeps duplicating the same skills and server configs across them.

Developer and ops teams maintaining internal AI tooling across a mixed set of clients.

Join the newsletter to get ahead with AI.

I share strategies, new modules, free guides, and tools to get you better at using AI in your work.