extract-mcp-servers-from-url

Instructions

Implementation Reference

• src/mcp_server_collector/server.py:66-117 (handler)

  Main tool handler function (@server.call_tool()). For 'extract-mcp-servers-from-url', fetches content from URL via helper, then extracts MCP servers using shared logic and returns as text.

  @server.call_tool()
  async def handle_call_tool(
      name: str, arguments: dict | None
  ) -> list[types.TextContent | types.ImageContent | types.EmbeddedResource]:
      if not arguments:
          raise ValueError("Missing arguments")
  
      content = None
      
      match name:
          case "extract-mcp-servers-from-url":
              url = arguments.get("url")
              if not url:
                  raise ValueError("Missing url")
  
              content = await call_fetch_tool(url)
              
          case "extract-mcp-servers-from-content":
              content = arguments.get("content")
              
          case "submit-mcp-server":
              url = arguments.get("url")
              avatar_url = arguments.get("avatar_url") or ""
              result = await submit_mcp_server(url, avatar_url)
              content = json.dumps(result)
  
              return [
                  types.TextContent(
                      type="text",
                      text=content,
                  )
              ]
          case _:
              raise ValueError(f"Unknown tool: {name}")
  
      if not content:
          raise ValueError("Missing content")
  
      logger.info(f"Fetched content from {url}: {content}")
  
      mcp_servers = await extract_mcp_servers_from_content(content)
      if not mcp_servers:
          raise ValueError("Extracted no MCP Servers")
  
      logger.info(f"Extracted MCP Servers from {url}: {mcp_servers}")
  
      return [
          types.TextContent(
              type="text",
              text=mcp_servers,
          )
      ]   

• src/mcp_server_collector/server.py:24-30 (schema)

  Input schema definition for 'extract-mcp-servers-from-url' tool: requires a 'url' string property.

  inputSchema={
      "type": "object",
      "properties": {
          "url": {"type": "string"},
      },
      "required": ["url"],
  },

• src/mcp_server_collector/server.py:20-31 (registration)

  Registration of the tool in @server.list_tools(): defines name, description, and input schema.

  return [
      types.Tool(
          name="extract-mcp-servers-from-url",
          description="Extract MCP Servers from a URL",
          inputSchema={
              "type": "object",
              "properties": {
                  "url": {"type": "string"},
              },
              "required": ["url"],
          },
      ),

• src/mcp_server_collector/extract.py:8-29 (helper)

  Helper function to extract MCP servers from content using OpenAI LLM in JSON object response format with custom prompt.

  async def extract_mcp_servers_from_content(content: str) -> str | None:
      client = OpenAI(
          api_key=os.getenv("OPENAI_API_KEY"),
          base_url=os.getenv("OPENAI_BASE_URL"),
      )
  
      user_content = extract_mcp_servers_prompt.format(content=content)
  
      logger.info(f"Extract prompt: {user_content}")
  
      chat_completion = client.chat.completions.create(
          messages=[
              {
                  "role": "user",
                  "content": user_content,
              }
          ],
          model=os.getenv("OPENAI_MODEL"),
          response_format={"type": "json_object"},
      )
  
      return chat_completion.choices[0].message.content

• src/mcp_server_collector/fetch.py:10-24 (helper)

  Helper function to fetch raw content from URL by invoking external 'mcp-server-fetch' via MCP client.

  async def call_fetch_tool(url: str):
      async with stdio_client(server_params) as (read, write):
          async with ClientSession(read, write) as session:
              await session.initialize()
  
              result = await session.call_tool(
                  "fetch",
                  arguments={
                      "url": url,
                      "max_length": 100000,
                      "raw": True
                  }
              )
  
              return result.content[0].text