> ## Documentation Index
> Fetch the complete documentation index at: https://docs.definable.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Getting Started with MCP

> Connect to your first MCP server and call tools from your agent.

This guide walks you through connecting to an MCP server, discovering its tools, and using them in an agent.

## Prerequisites

* An MCP server to connect to (we'll use the filesystem server from npm)
* Node.js installed (for `npx`)

## Step 1: Define the Server

Create a configuration for the MCP server:

```python theme={null}
from definable.mcp import MCPConfig, MCPServerConfig

config = MCPConfig(servers=[
    MCPServerConfig(
        name="filesystem",
        transport="stdio",
        command="npx",
        args=["-y", "@modelcontextprotocol/server-filesystem", "/tmp"],
    ),
])
```

## Step 2: Create the Toolkit

Wrap the configuration in an `MCPToolkit`:

```python theme={null}
from definable.mcp import MCPToolkit

toolkit = MCPToolkit(config=config)
```

## Step 3: Connect and Use

Use the toolkit as a context manager to manage the server connection lifecycle:

```python theme={null}
from definable.agent import Agent
from definable.model import OpenAIChat

async with toolkit:
    # Inspect available tools
    for t in toolkit.tools:
        print(f"  {t.name}: {t.description}")

    # Create an agent with MCP tools
    agent = Agent(
        model=OpenAIChat(id="gpt-4o"),
        toolkits=[toolkit],
        instructions="You can read and write files. Help the user with file operations.",
    )

    output = await agent.arun("Create a file called hello.txt with 'Hello, MCP!' in /tmp")
    print(output.content)
```

## Complete Example

```python theme={null}
import asyncio
from definable.agent import Agent
from definable.mcp import MCPToolkit, MCPConfig, MCPServerConfig
from definable.model import OpenAIChat

async def main():
    config = MCPConfig(servers=[
        MCPServerConfig(
            name="filesystem",
            transport="stdio",
            command="npx",
            args=["-y", "@modelcontextprotocol/server-filesystem", "/tmp"],
        ),
    ])

    async with MCPToolkit(config=config) as toolkit:
        agent = Agent(
            model=OpenAIChat(id="gpt-4o"),
            toolkits=[toolkit],
        )
        output = await agent.arun("List the files in /tmp and tell me what's there.")
        print(output.content)

asyncio.run(main())
```

## Using SSE or HTTP Transports

For remote servers, use SSE or HTTP:

<CodeGroup>
  ```python SSE theme={null}
  MCPServerConfig(
      name="remote",
      transport="sse",
      url="https://mcp-server.example.com/sse",
  )
  ```

  ```python HTTP theme={null}
  MCPServerConfig(
      name="remote",
      transport="http",
      url="https://mcp-server.example.com/mcp",
  )
  ```
</CodeGroup>

## Sync Usage

If you're not in an async context, use the sync wrappers:

```python theme={null}
with MCPToolkit(config=config) as toolkit:
    agent = Agent(model=OpenAIChat(id="gpt-4o"), toolkits=[toolkit])
    output = agent.run("List files in /tmp")
    print(output.content)
```

## Next Steps

* [Configuration](/mcp/configuration) — Timeouts, headers, tool filtering
* [Resources](/mcp/resources) — Read data from MCP servers
* [Prompts](/mcp/prompts) — Use prompt templates