vibewatch
Stop copy-pasting terminal errors to ChatGPT/Claude. VibeWatch wraps your dev commands and lets AI assistants see your terminal in real-time via Model Context Protocol (MCP).
claude mcp add --transport stdio krjordan-vibewatch npx vibewatch npm run dev
How to use
VibeWatch is a CLI tool that sits between your development process and Claude (or any MCP-compatible AI assistant) to stream terminal output in real time. It monitors the commands you run and exposes the resulting terminal activity to your MCP client so the AI can provide context, debugging help, or proactive suggestions without you having to paste logs. The MCP integration exposes a small set of tools that you can query to inspect recent terminal output, crash context, and errors, as well as test connectivity to the local API.
Available MCP tools include get_terminal_output for retrieving recent terminal lines with optional filtering, get_crash_context for detailed crash data with stack traces, get_recent_errors for a quick view of recent issues, and ping to verify connectivity. These tools are designed to help Claude understand the state of your development environment and assist with debugging by providing structured, machine-readable data about your terminal activity and errors.
How to install
Prerequisites:
- Node.js and npm installed on your system (node -v and npm -v).
- Access to the command line/terminal.
Installation options:
Option 1: Global install
npm install -g vibewatch
vibewatch npm run dev
Option 2: Run with npx (no install)
npx vibewatch npm run dev
Option 3: Local development (from repository)
git clone https://github.com/krjordan/vibewatch.git
cd vibewatch
npm install
npm run build
node dist/cli.js npm run dev
After installation, you should be able to wrap commands with vibewatch to stream terminal output to the MCP client.
Additional notes
Tips and notes:
- The default API server port is 3333 (you can override with --port).
- The log buffer size defaults to 100 lines; adjust with --buffer-size for longer-running processes.
- Enable verbose stack traces with --verbose to include node_modules in stack traces.
- Use --raw to disable noise filtering when debugging (e.g., to see every line).
- Keep-alive behavior after a crash is controlled with --keep-alive (seconds).
- The MCP integration exposes several tools (get_terminal_output, get_crash_context, get_recent_errors, ping). Make sure your MCP client is configured to access the local API if you’re using Claude Desktop/Cursor settings.
- If you encounter port conflicts, ensure the specified port is open and not used by another service.
Related MCP Servers
gemini -tool
MCP server that enables AI assistants to interact with Google Gemini CLI, leveraging Gemini's massive token window for large file analysis and codebase understanding
claude-code-open
Open source AI coding platform with Web IDE, multi-agent system, 37+ tools, MCP protocol. MIT licensed.
mobile
A Model Context Protocol (MCP) server that provides mobile automation capabilities.
Email MCP server with full IMAP + SMTP support — read, search, send, manage, and organize email from any AI assistant via the Model Context Protocol
classmcp
MCP server for AI-assisted CSS development. 77% token savings with semantic class patterns. Supports Tailwind, Bootstrap, UnoCSS, Tachyons.
wordpress-boost
MCP server exposing WordPress internals to AI - hooks, schema, security, and test data. Inspired by Laravel Boost.