🤖 MCP Server for Copilot Studio Agents

Name: 🤖 MCP Server for Copilot Studio Agents
Author: CrewAakash

@CrewAakash

Repository

About 🤖 MCP Server for Copilot Studio Agents

MCP server wrapper for the agents built using MSFT Copilot studio

#microsoft #mcs #copilot

Basic information

Config

Add this server to your MCP-compatible client using the configuration below.

{
  "mcpServers": {
    "agent-name": {
      "command": "uv",
      "args": [
        "--directory",
        "<PATH_TO_THE_PARENT_FOLDER>",
        "run",
        "mcp",
        "run",
        "<PATH_TO_THE_PARENT_FOLDER>/src/main.py"
      ],
      "env": {
        "DIRECTLINE_ENDPOINT": "endpoint-url",
        "COPILOT_AGENT_SECRET": "secret-key"
      }
    }
  }
}

Tools

No tools detected

We auto-extract tools from the README. The maintainer can list them under a ## Tools heading to populate this section.

Overview

What is 🤖 MCP Server for Copilot Studio Agents?

An MCP server that connects to Microsoft Copilot Studio agents via the DirectLine API, enabling any MCP-compatible client to send queries and receive responses while maintaining conversation context.

How to use 🤖 MCP Server for Copilot Studio Agents?

Install with uv or pip, set the required environment variables DIRECTLINE_ENDPOINT and COPILOT_AGENT_SECRET, configure agent definitions in src/main.py, then run the server with mcp dev src/main.py or via a client like Claude Desktop using the provided JSON configuration.

Key features of 🤖 MCP Server for Copilot Studio Agents

Connects to Copilot Studio agents through DirectLine API
Maintains conversation context across multiple queries
Provides a single query_agent tool for interaction
Supports stateful conversations with conversation ID and watermark tracking
Configurable agent definitions with name and description
Works with any MCP-compatible client

Use cases of 🤖 MCP Server for Copilot Studio Agents

Send queries to a Copilot Studio agent and receive structured responses
Build multi-turn conversations with persistent context
Integrate Copilot Studio agents into AI assistants like Claude Desktop
Test and debug agent interactions using MCP Inspector

FAQ from 🤖 MCP Server for Copilot Studio Agents

What does this server do?

It exposes a Copilot Studio agent as an MCP tool, allowing any MCP client to query the agent and maintain conversation history.

What are the prerequisites to use this server?

You need Python 3.12+, a Copilot Studio agent with Direct Line channel security configured, and the environment variables DIRECTLINE_ENDPOINT and COPILOT_AGENT_SECRET.

What tools does the server expose?

It exposes one tool: query_agent, which sends a query to the Copilot Studio agent and returns a structured response with success/error status.

How do I use it with Claude Desktop?

Add a JSON entry to claude_desktop_config.json under mcpServers, specifying the command (uv, python, or docker) and the required environment variables.

Does the server store any data or state?

Conversation state (conversation IDs and watermarks) is maintained in memory across queries for the duration of the session; no persistent storage is described.