auto-mcp

Automatically convert functions, tools and agents to MCP servers.

🧩 Installing auto-mcp

You can install the SDK using PyPI or from source.

Install within an existing project

If you want to install auto-mcp as part of an existing project (e.g. to automatically convert existing agents or tools to MCP servers), it is good practice to do so within a dedicated virtual environment.

1. (Optional) Create a new virtual environment

If you don't already have a virtual environment, create a new one using uv:

uv init
source .venv/bin/activate

2. Add auto-mcp to your dependencies

Then add auto-mcp. In uv this looks like:

uv add auto-mcp

If not using uv or poetry, you can also use pip:

pip install auto-mcp

3. Initialize automcp

Use the CLI to run the following command:

automcp init

This will create an automcp.py file at the root of your project.

4. Modify the automcp.py file

Modify the automcp.py file to:

1. Import the agents or tools you would like to convert from your existing project
2. Import the adapters from automcp that correspond to the agent framework that you are using
3. Define an input schema for the agent or tool

A simple example automcp.py for a CrewAI agent might look like:

from marketing_posts.crew import MarketingPostsCrew
from automcp import crewai_adapter

class InputSchema(BaseModel):
    customer_domain: str
    project_description: str

mcp = FastMCP("my MCP Server")
name = "Marketing Crew"
description = "A crew that creates marketing posts"
input_schema = InputSchema

tool = crewai_adapter(
    crewai_class=MarketingPostsCrew,
    name=name,
    description=description,
    input_schema=input_schema,
)
mcp.add_tool(
    tool,
    name=name,
    description=description,
)

if __name__ == "__main__":
    serve_stdio(mcp)  # Launch the MCP server 

5. Configure your .env file

Add any required environmental variables:

OPENAI_API_KEY=<your_openai_api_key>
SERPER_API_KEY=<your_serper_api_key>

6. Start the Server(s)

Using STDIO:

uv run serve_stdio

Using SSE:

uv run serve_sse

7. Testing and Integration

Cursor

Here is an example configuration of mcp.json that runs the MCP server using STDIO:

{
   "mcpServers": {
       "Marketing Crew": {
           "command": "uvx",
           "args": [
               "--from",
               "git+https://github.com/K-Mistele/example-mcp serve_stdio"
           ],
           "env": {
               "OPENAI_API_KEY": "...",
               "SERPER_API_KEY": "..."
           }
       }
   }
}

With SSE:

{
   "mcpServers": {
       "Marketing Crew": {
           "url": "http://localhost:8000/sse"
       }
   }
}

8. Publish

Coming soon!

Install from source

If you are a developer contributing to AutoMCP, you will want to install from source using:

git clone https://github.com/NapthaAI/auto-mcp.git
cd auto-mcp
uv venv
source .venv/bin/activate
uv pip install .