MCP Server

The Model Context Protocol (MCP) server allows AI tools and LLMs to interact withRepo Cat programmatically.

Introduction

The Model Context Protocol (MCP) is an open standard that enables AI assistants to interact with external systems.Repo Cat provides an MCP server that exposes tools, resources, and prompts for repository analysis.

With the MCP server, AI tools like Claude, GPT, and others can trigger scans, retrieve results, manage settings, and more — all through a standardized protocol.

Server Name: repo-cat-mcp
Version: 1.0.0
Transport: Streamable HTTP (Stateless)

Authentication

The MCP server uses the same API key authentication as the REST API. You must include a valid API key with each request.

Authentication Methods

Header: Authorization: Bearer <api-key>
Query Parameter: ?apiKey=<api-key>

Stateless Mode

The MCP server operates in stateless mode. Each request creates a new transport instance with no session persistence. The user context (userId) is extracted from the API key and passed to tools via the request context.

Endpoint

The MCP server is available at the following endpoint:

http://repo-cat.futoro.co.uk/mcp

Transport Details

Property	Value
Transport	Streamable HTTP
Session	Stateless (no session persistence)
Method	POST (Streamable HTTP)
Content-Type	application/json

Tools

The MCP server exposes 9 tools for interacting with Repo Cat. All tools require authentication.

trigger_scan

Trigger a new repository scan (async)

Parameter	Type	Description
repo	string	GitHub repository in format "owner/repo"
level	string	"basic", "standard", or "deep"
reference	string	Branch or tag to scan (optional)

get_scan

Get scan details by ID

Parameter	Type	Description
scanId	number	The ID of the scan to retrieve

list_scans

List user's scans with filtering

Parameter	Type	Description
page	number	Page number for pagination
status	string	Filter by status (optional)
level	string	Filter by scan level (optional)

cancel_scan

Cancel an in-progress scan

Parameter	Type	Description
scanId	number	The ID of the scan to cancel

delete_scan

Delete a scan by ID

Parameter	Type	Description
scanId	number	The ID of the scan to delete

get_scan_pdf

Generate and retrieve PDF report (returns base64-encoded PDF)

Parameter	Type	Description
scanId	number	The ID of the scan for PDF generation

get_user_info

Get current user details and subscription

No parameters required

get_settings

Retrieve user settings

No parameters required

update_settings

Update user settings

Parameter	Type	Description
email_notifications_enabled	boolean	Enable email notifications (optional)
email_on_scan_started	boolean	Receive email when scan starts (optional)
email_on_scan_completed	boolean	Receive email when scan completes (optional)
email_on_scan_failed	boolean	Receive email when scan fails (optional)
email_on_scan_cancelled	boolean	Receive email when scan is cancelled (optional)
webhook_enabled	boolean	Enable webhook notifications (optional)
webhook_url	string	Webhook URL for notifications (optional)
webhook_on_scan_started	boolean	Trigger webhook when scan starts (optional)
webhook_on_scan_completed	boolean	Trigger webhook when scan completes (optional)
webhook_on_scan_failed	boolean	Trigger webhook when scan fails (optional)
webhook_on_scan_cancelled	boolean	Trigger webhook when scan is cancelled (optional)

Resources

Resources provide access to data entities. The MCP server exposes one resource for accessing scan results.

URI	Description
scan://{scanId}	Full scan result with metadata, heuristics, and analysis

Example: Accessing a Scan Resource

// Access scan resource via MCP
const result = await client.readResource({
  uri: 'scan://123',
});

console.log(result.contents[0].text);

Prompts

Prompts are guided templates that help AI tools perform complex tasks. The MCP server provides a prompt for analyzing repositories.

analyze-repo

Guided prompt template that helps AI tools analyze a GitHub repository by triggering a scan and polling for results.

When an AI tool uses this prompt, it guides the assistant through:

Calling the trigger_scan tool to start a scan
Polling the get_scan tool to check progress
Providing a comprehensive summary once complete

Example Usage

Connect to the MCP server using the official MCP SDK.

JavaScript: Connecting with MCP SDK

import { Client } from '@modelcontextprotocol/sdk/client/index.js';
import { StreamableHTTPClientTransport } from '@modelcontextprotocol/sdk/client/streamableHttp.js';

const transport = new StreamableHTTPClientTransport(
  new URL('http://repo-cat.futoro.co.uk/mcp')
);

const client = new Client(
  { name: 'example-client', version: '1.0.0' },
  { capabilities: {} }
);

await client.connect(transport);

// Call a tool
const result = await client.callTool({
  name: 'trigger_scan',
  arguments: {
    repo: 'facebook/react',
    level: 'standard',
  },
});

console.log(result);

cURL: Calling a Tool Directly

You can also call MCP tools directly using cURL with JSON-RPC:

curl -X POST "http://repo-cat.futoro.co.uk/mcp" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer your-api-key-here" \
  -d '{
    "jsonrpc": "2.0",
    "id": 1,
    "method": "tools/call",
    "params": {
      "name": "trigger_scan",
      "arguments": {
        "repo": "facebook/react",
        "level": "standard"
      }
    }
  }'

Use Case: AI Repository Analysis

An AI tool can use the analyze-repo prompt to analyze a repository:

AI loads the analyze-repo prompt
Prompt guides AI to call trigger_scan with repo details
AI polls get_scan to check scan status
Once complete, AI reads the scan://{scanId} resource
AI provides a comprehensive analysis to the user

API Reference - For REST API documentation instead of MCP