> ## 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.

# MCP Overview

> Connect to MCP servers and use their tools, resources, and prompts.

The **Model Context Protocol (MCP)** is an open standard for connecting AI agents to external tools and data sources. Definable's MCP module lets you connect to any MCP-compatible server and use its capabilities inside your agents.

## What MCP Provides

MCP servers expose three types of capabilities:

| Capability    | Description                  | Example                                      |
| ------------- | ---------------------------- | -------------------------------------------- |
| **Tools**     | Functions the agent can call | File operations, database queries, API calls |
| **Resources** | Data the agent can read      | Files, configurations, live data feeds       |
| **Prompts**   | Reusable prompt templates    | Analysis templates, code review prompts      |

## Architecture

```mermaid theme={null}
flowchart TD
  Agent --> MCPToolkit --> MCPClient
  MCPClient -->|JSON-RPC| stdio
  MCPClient -->|JSON-RPC| SSE
  MCPClient -->|JSON-RPC| HTTP
  stdio --> FSSrvr["FS Server"]
  SSE --> GitSrvr["Git Server"]
  HTTP --> APISrvr["API Server"]
```

## Transport Types

Definable supports all three MCP transport types:

| Transport | Protocol                  | Best For                      |
| --------- | ------------------------- | ----------------------------- |
| **stdio** | Subprocess stdin/stdout   | Local CLI tools, npm packages |
| **SSE**   | HTTP + Server-Sent Events | Remote servers, legacy MCP    |
| **HTTP**  | Streamable HTTP           | Remote servers, modern MCP    |

## Quick Example

Connect to a filesystem MCP server and use it in an agent:

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

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 all files in /tmp")
    print(output.content)
```

## Key Components

<CardGroup cols={2}>
  <Card title="Getting Started" icon="rocket" href="/mcp/getting-started">
    Connect to your first MCP server step by step.
  </Card>

  <Card title="Configuration" icon="gear" href="/mcp/configuration">
    Configure servers, transports, timeouts, and tool filtering.
  </Card>

  <Card title="Resources" icon="folder-open" href="/mcp/resources">
    Read data from MCP resource providers.
  </Card>

  <Card title="Prompts" icon="message" href="/mcp/prompts">
    Use reusable prompt templates from MCP servers.
  </Card>

  <Card title="Error Handling" icon="triangle-exclamation" href="/mcp/error-handling">
    Handle connection failures, timeouts, and reconnection.
  </Card>

  <Card title="Mock Servers" icon="flask-vial" href="/mcp/mock-servers">
    Test MCP integrations without real servers.
  </Card>
</CardGroup>