OpenAI, Microsoft, Google, and Amazon just agreed on one AI agent plugin format
Questions this post answers
What is the Agent Plugins 1.0.0 spec and what does its directory structure look like?
Agent Plugins 1.0.0 is a vendor-neutral open spec for packaging AI agent skills and MCP servers into portable plugins, co-authored by OpenAI, Amazon, Microsoft, Cursor, and Vercel, with Google as a Core Maintainer. The format is minimal: a directory containing a plugin.json, a skills/ folder, and an optional mcp.json. Install mechanisms, distribution protocols, and permission models are intentionally left to each client. Developers building or evaluating AI agent integrations track spec developments like this on daily.dev.
What is the difference between agent skills and MCP servers in the Agent Plugins spec?
Agent skills are packaged platform knowledge — for example, how to deploy to Railway or query Supabase — while MCP servers give agents live account access to real infrastructure. Plugins bundle both. The practical guidance is to start with read-only skills before granting an agent MCP access to production systems, as a basic safety measure when autonomous tools touch real accounts. Teams deciding how much access to grant AI agents in production find the tradeoffs discussed on daily.dev.