Agent Plugins
A portable package format for reusable components that extend AI agents.
Agent Plugins is an open, vendor-neutral specification for packaging reusable components that extend AI agents into distributable plugins. Its portable contract builds on two existing standards: Agent Skills and the Model Context Protocol.
The portable package
An Agent Plugin is a directory with a required manifest and optional components in fixed locations:
my-plugin/
├── plugin.json
├── skills/
│ └── summarize/
│ ├── SKILL.md
│ ├── scripts/
│ └── references/
├── mcp.json
└── com.example.client/
└── hooks/plugin.jsonidentifies the plugin and the Agent Plugins version it targets.skills/contains Agent Skills in the format defined by the Agent Skills specification.mcp.jsondescribes stdio, Streamable HTTP, or legacy HTTP+SSE MCP servers.- Reverse-domain extension namespaces let individual clients add behavior without changing the portable core.
Agent Plugins intentionally leaves distribution, marketplaces, installation workflows, permissions, and client UX outside the portable contract.
Choose your path
Build a plugin
Create a portable package with skills and MCP servers.
Implement a client
Load, validate, and run Agent Plugins safely.
Read the specification
Consult the complete normative specification.
Browse the schemas
Use the canonical plugin and MCP JSON Schemas.
Source and governance
The versioned specification is authoritative. Project governance, contribution guidance, schemas, and release history live in the Agent Plugins specification repository.