Our MCP dev kit

We made a simple, clean and pythonic implementation of MCP protocol - MCPager to leverage the power of MCP on our platform. MCPager acts like glue between an AI agent and any open MCP server, e.g. any one listed in the Docker MCP collection..

It also provides an out-of-the-box connector to LangGraph, so you can easily build an integration with your own AI agent using one of the most popular open agentic frameworks.

Integration with an MCP server of your choice

To use our dev kit with an AI agent, you need an MCP server first. You can use any open MCP server or build your own. Refer to the examples of basic integrations with MCPager and LangChain/LangGraph.

For example, you can connect to a server based on FastMCP framework using MCPager:

fastmcp serve examples/echo_server.py --transport http

Basic example

This is how you can initialize an MCP client with LangGraph backend and call tools:

from mcpager.client import StdioTransport, HttpTransport, Session, MCPClient
from mcpager.adapters import LangGraphBackend

transport = HttpTransport('http://localhost:8000/mcp')
# or you can use StdioTransport for local testing like this:
# transport = StdioTransport(["fastmcp", "run", "echo_server.py"])
session = Session(transport)
backend = LangGraphBackend()
client = MCPClient(backend, session)

client.initialize()

tools = client.list_tools()
print(tools)

result = client.call_tool("echo", text="Hello from MCPager!")
print(result)