# Model Context Protocol (MCP)

The PureCode AI Extension has significantly enhanced its capabilities by introducing support for the Model Context Protocol (MCP). This integration allows users to connect to a diverse range of backend services, which can include local binaries, applications running within Docker containers, or remote HTTP services.

> Connect your PureCode AI Extension to diverse backends—local binaries, Docker containers, and remote services—using a simple, extensible JSON configuration format.

<figure><img src="https://1934073637-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FVx22AmjG9CHfbdGPtjJI%2Fuploads%2F2t16xJ1FQX3ID3uvLRdv%2FScreenshot%202025-07-24%20at%207.01.30%E2%80%AFPM.png?alt=media&#x26;token=fb6036c9-153d-4063-be52-b24393bc82a3" alt=""><figcaption></figcaption></figure>

<table data-view="cards"><thead><tr><th></th><th></th></tr></thead><tbody><tr><td><strong>Flexible Configuration</strong></td><td>JSON-based format supporting multiple servers with stdio and SSE transport protocols.</td></tr><tr><td><strong>Automatic Discovery</strong></td><td>Tools are automatically detected and made available in the sidebar with real-time status.</td></tr><tr><td>Extensive Architecture</td><td>Add new MCP servers without code changes—simply update your configuration.</td></tr></tbody></table>

### What is MCP?

The **Model Context Protocol (MCP)** is an open protocol designed to standardize how applications provide context, data sources, and tools to Large Language Models (LLMs). Think of MCP as a universal adapter, much like a USB-C port for AI applications.&#x20;

### Key Benefits

<table data-card-size="large" data-view="cards"><thead><tr><th></th><th></th></tr></thead><tbody><tr><td><strong>Plug-and-Play Extensibility</strong></td><td>Users can effortlessly add, remove, or swap out MCP servers by making simple changes to a configuration file or through the user interface. New MCP-compatible tools can be integrated without requiring updates to the core extension.</td></tr><tr><td><strong>Unified User Experience</strong></td><td>Regardless of the underlying backend, all tools exposed via MCP servers appear and behave consistently within the PureCode AI Extension's UI, reducing learning curves and streamlining workflows.</td></tr></tbody></table>

### Configuring MCP Servers

Configuring MCP servers involves defining them in a structured JSON format. The configuration file serves as the central point for managing all MCP servers accessible to the PureCode AI Extension.

Basic Configuration Structure :

{% code overflow="wrap" %}

```json
{
  "mcpServers": {
    "your-server-id": {
      "name": "Server Name",
      "id": "your-server-id",
      "transport": {
        "type": "stdio|sse",
        // For stdio transport:
        "command": "npx|docker|node",
        "args": ["-y", "package-name", "--optional-flag"],
        "env": {
          "API_KEY": "your_api_key"
        }
        // For SSE transport:
        // "url": "https://your-server.com/sse"
      },
      "enabled": true
    }
  }
}
```

{% endcode %}

### Transport Mechanisms

<table data-card-size="large" data-view="cards"><thead><tr><th></th><th></th><th></th></tr></thead><tbody><tr><td><strong>stdio (Standard Input/Output)</strong></td><td>Used for MCP servers launched as local processes. The extension starts the process and communicates via stdin/stdout.</td><td><strong>Fields:</strong> command, args, env (optional)</td></tr><tr><td>sse (Server-Sent Events)</td><td>Used for MCP servers exposing HTTP endpoints. The extension connects to the URL and receives real-time updates.</td><td><strong>Field:</strong> url</td></tr></tbody></table>

{% hint style="info" %}
**Securing Credentials**

Use the `env` object to securely pass API keys and tokens without hardcoding them in configuration files. This prevents accidental exposure in version control systems.
{% endhint %}

### Key Configuration Fields

| Field     | Description                           | Required | Example                  |
| --------- | ------------------------------------- | -------- | ------------------------ |
| name      | Human-readable name for display in UI | Yes      | "Filesystem Indexer"     |
| id        | Unique identifier for the server      | Yes      | "file-system"            |
| transport | Connection protocol and parameters    | Yes      | { "type": "stdio", ... } |

### Technical Details and Advanced Features

<table data-card-size="large" data-view="cards"><thead><tr><th></th><th></th></tr></thead><tbody><tr><td><strong>Automatic Discovery</strong></td><td>The extension automatically parses the mcpServers configuration, establishes connections, and queries each server for available tools. This happens transparently during startup or configuration reload.</td></tr><tr><td><strong>Extensible Transport Layer</strong></td><td>The architecture supports multiple transport mechanisms (stdio, SSE) and is designed to accommodate future protocols without fundamental changes to the core extension.</td></tr><tr><td><strong>Seamless Integration</strong></td><td>MCP-enabled tools are indistinguishable from native tools in the user interface. The AI can autonomously invoke these tools when appropriate to fulfill user requests.</td></tr><tr><td><strong>Deployment Flexibility</strong></td><td>Comprehensive support for local processes, Docker containers, and remote HTTP services ensures users can choose the deployment model that best fits their needs.</td></tr></tbody></table>

### Supported Deployment Models

<table data-view="cards"><thead><tr><th></th><th></th></tr></thead><tbody><tr><td><strong>Local Processes</strong></td><td>Run directly on your machine using stdio transport</td></tr><tr><td><strong>Docker Containers</strong></td><td>Isolated, consistent execution environments</td></tr><tr><td><strong>Remote Services</strong></td><td>Connect to HTTP endpoints via SSE transport</td></tr></tbody></table>