MCP Server

vibecop includes a built-in MCP (Model Context Protocol) server that exposes scan, check, and explain tools over stdio transport. This enables integration with any MCP-compatible AI coding tool.

Starting the Server

vibecop serve

The server starts on stdio (standard input/output) and logs status messages to stderr. It runs until terminated with SIGINT or SIGTERM.

MCP Client Configuration

Configure your MCP client to connect to vibecop:

{
  "mcpServers": {
    "vibecop": {
      "command": "npx",
      "args": ["vibecop", "serve"]
    }
  }
}

Available Tools

The MCP server exposes three tools:

vibecop_scan

Scan a directory for AI code quality issues.

Parameters:

Parameter	Type	Required	Default	Description
path	string	No	Current working directory	Directory to scan
maxFindings	number	No	50	Maximum findings to return

Response: JSON object with findings, filesScanned, and errors arrays.

vibecop_check

Check a single file for AI code quality issues.

Parameters:

Parameter	Type	Required	Default	Description
file_path	string	Yes	—	Absolute or relative path to the file
maxFindings	number	No	50	Maximum findings to return

Response: JSON object with findings, filesScanned, and errors arrays.

vibecop_explain

Get detailed information about what a specific detector checks for.

Parameters:

Parameter	Type	Required	Description
detector_id	string	Yes	The detector ID (e.g., unsafe-shell-exec, god-function)

Response: JSON object with id, name, description, severity, category, and languages.

If the detector ID is not found, returns an error with the list of all available detector IDs.

Tool-Specific Setup

Continue.dev

Add to .continue/config.json:

{
  "mcpServers": [
    {
      "name": "vibecop",
      "command": "npx",
      "args": ["vibecop", "serve"]
    }
  ]
}

Amazon Q

Configure the MCP server in your Amazon Q developer settings:

{
  "mcpServers": {
    "vibecop": {
      "command": "npx",
      "args": ["vibecop", "serve"]
    }
  }
}

Zed

Add to your Zed MCP settings:

{
  "mcpServers": {
    "vibecop": {
      "command": "npx",
      "args": ["vibecop", "serve"]
    }
  }
}

Architecture

The MCP server is built on the @modelcontextprotocol/sdk package and uses stdio transport (JSON-RPC over stdin/stdout). It reuses the same scan() and checkFile() engine functions as the CLI.

MCP Client (Continue.dev / Amazon Q / Zed)
  ↕ JSON-RPC over stdio
vibecop MCP Server
  → vibecop_scan  → engine.scan()
  → vibecop_check → engine.checkFile()
  → vibecop_explain → detector metadata lookup

The server handles graceful shutdown on SIGINT and SIGTERM. All logging goes to stderr so it does not interfere with the MCP protocol on stdout.

When to Use MCP vs Hooks

Criteria	MCP Server (Tier 3)	Hooks (Tier 1)
Tool support	Any MCP client	Tool-specific hooks
Invocation	On-demand by agent	Automatic on file edit
Blocking	Agent-controlled	Deterministic (exit code)
Best for	MCP-native tools	Claude Code, Cursor

Use the MCP server when your tool supports MCP natively (Continue.dev, Amazon Q, Zed). Use hooks when your tool supports them (Claude Code, Cursor, Codex CLI) — hooks provide stronger guarantees because they execute deterministically.