MCP 服务器

aimarket-mcp-packager

Model Context Protocol (MCP) stdio server built with the official MCP Python SDK (FastMCP). Exposes 3 MCP tools: package_capability, generate_dockerfile, generate_claude_desktop_config — package AIMarket capabilities as Docker images + MCP manifest + Claude Desktop config. Entrypoint: mcp_stdio_server.py; root Dockerfile + glama.json in repo.

README

aimarket-mcp-packager — MCP server

This repository ships a Model Context Protocol (MCP) server that packages AIMarket capabilities into self-hosted MCP server products (Docker image metadata, MCP manifest, Claude Desktop config).

Transport: stdio (mcp_stdio_server.py). Built with the official MCP Python SDK (mcp package, FastMCP).

Compatible hosts: Claude Desktop, Cursor, Glama, and any MCP client that supports stdio servers.

MCP server in this repo

Item	Location
MCP entrypoint	`mcp_stdio_server.py`
MCP SDK dependency	`requirements-mcp.txt` — `mcp>=1.6,<4`
Glama / Docker run	`Dockerfile`, `glama.json`
Tool implementation	`aimarket_mcp_packager/packager_core.py`

Run locally:

pip install -r requirements-mcp.txt
pip install -e .
python mcp_stdio_server.py

Or with Docker (stdio MCP server):

docker build -t aimarket-mcp-packager .
docker run --rm -i aimarket-mcp-packager

MCP tools

The server exposes 3 tools (no MCP resources or prompts):

Tool	Description
`package_capability`	Build an MCP server package: Docker image name, `mcp_manifest`, subscription tiers, connection string
`generate_dockerfile`	Generate a Dockerfile template for the packaged MCP server
`generate_claude_desktop_config`	Generate a `claude_desktop_config.json` snippet (`mcpServers`) for docker-run MCP

Example tool call (conceptual):

{
  "name": "package_capability",
  "arguments": {
    "capability_id": "translate.multi@v2",
    "product_id": "prod-001",
    "name": "Lyra",
    "description": "Multilingual translator",
    "input_schema_json": "{\"type\":\"object\",\"properties\":{\"text\":{\"type\":\"string\"}}}"
  }
}

MCP resources

This server does not register MCP resources or prompts — only the tools above. Output is JSON text returned from each tool call.

Client configuration

Claude Desktop

{
  "mcpServers": {
    "aimarket-mcp-packager": {
      "command": "python",
      "args": ["/path/to/aimarket-mcp-packager/mcp_stdio_server.py"]
    }
  }
}

Docker variant (see generate_claude_desktop_config tool output):

{
  "mcpServers": {
    "aimarket-mcp-packager": {
      "command": "docker",
      "args": ["run", "--rm", "-i", "aimarket-mcp-packager:latest"]
    }
  }
}

Cursor

Add the same stdio server under Settings → MCP (command + args pointing at mcp_stdio_server.py or the Docker image above).

Python library (optional)

The same packaging logic is available as a library and as an AIMarket Hub plugin (pip install aimarket-mcp-packager). Hub HTTP routes are optional; the MCP server runs standalone via mcp_stdio_server.py.

from aimarket_mcp_packager.mcp_packager import MCPPackager

packager = MCPPackager()
pkg = packager.package(
    capability_id="translate.multi@v2",
    product_id="prod-001",
    name="Lyra",
    description="Multilingual translator",
    input_schema={"type": "object", "properties": {"text": {"type": "string"}}},
    price_per_call_usd=0.40,
)

print(pkg.docker_image)
print(pkg.mcp_manifest)
print(packager.generate_claude_desktop_config(pkg))