Packaging as an Agent Plugin (v1.0.0)

Once your Go MCP server is operational, package it as an Agent Plugin so assistants (Cursor, OpenCode, Claude Code) can discover both the MCP server and its accompanying Agent Skills in a single bundle.

---

1. Package Structure

An Agent Plugin bundle pairs your MCP server configuration (mcp.json) with declarative skills (skills/):

my-mcp-plugin/
├── plugin.json               # Plugin root manifest (name, version, metadata)
├── mcp.json                  # MCP server connection definitions
├── skills/
│   └── domain-analysis/
│       ├── SKILL.md          # Agent instructions & prompt workflows
│       └── references/
│           └── playbook.md   # Domain reference guides
├── LICENSE
└── README.md

---

2. Manifest Configuration (plugin.json)

Create plugin.json at the root, targeting the canonical v1.0.0 schema:

{
  "$schema": "https://agent-plugins.org/schemas/1.0.0/plugin.schema.json",
  "name": "my-mcp-plugin",
  "version": "1.0.0",
  "description": "Production Go MCP tools and workflow skills",
  "author": {
    "name": "Your Name",
    "url": "https://github.com/username"
  },
  "repository": "https://github.com/username/my-mcp-plugin",
  "license": "Apache-2.0",
  "keywords": ["mcp", "go", "ai-agents"]
}

---

3. Server Configuration (mcp.json)

Declare your remote Streamable HTTP or local stdio endpoints in mcp.json:

{
  "$schema": "https://agent-plugins.org/schemas/1.0.0/mcp.schema.json",
  "mcpServers": {
    "remote-service": {
      "type": "streamable-http",
      "url": "https://your-server.run.app/sse"
    }
  }
}

---

4. Authoring Accompanying Skills

Add skills/<skill-name>/SKILL.md documents providing instructions to AI agents on how and when to orchestrate your tools.

For advanced multi-plugin authoring, marketplace manifest generation, and automated validation scripts, refer to the agent-plugin-authoring toolchain.