mcpconnect

MCP Connect provides a browser-based development environment that lets you connect to any MCP server (via HTTP, WebSocket, or SSE), test tools using Claude or OpenAI models, and inspect protocol traffic. It also lets you enable or disable specific tools to manage token usage and export conversations in JSON or Markdown. This makes it ideal for iterating on MCP tool development, validating protocol behavior, and debugging integration with external MCP servers. To get started, run the Quick Start command to launch the MCP Connect client in your browser, then use the interface to connect to an MCP server, pick a tool to run, and observe the protocol messages being exchanged in real time.

Prerequisites:

• Node.js and npm installed on your machine
• Internet access to fetch packages

Step-by-step installation:

1. Ensure Node.js and npm are installed. You can verify with: node -v npm -v

2. Install or run MCP Connect via npx (no global install required): npx @mcpconnect/cli@latest

3. When prompted, the CLI will start a browser-based UI that connects to MCP servers. If you prefer to pin a specific version, replace @latest with a version tag, e.g. @1.2.3.

Optional: If you want to install locally for ongoing development, you can clone the repository and install dependencies, then run the local dev script as defined by the project maintainers.

Notes:

• You may need to allow your browser to open a local address (e.g., http://localhost:3000) depending on your environment.
• Ensure network access to the MCP server endpoints you intend to connect to.

Tips and common issues:

• If a connected MCP server uses a non-standard port or requires authentication, configure the connection settings in the MCP Connect UI.
• Some tools may require API keys or access tokens (e.g., OpenAI, Claude). Keep these credentials secure and consider using environment variables or secret managers if supported by your setup.
• Enable or disable tools on the Tools page to control token usage and data exported with conversations.
• If you encounter browser script errors, try reloading the page or clearing the cache. Ensure your browser supports WebSocket and SSE as required by MCP Connect.
• When exporting conversations, choose the desired format (JSON or Markdown) for downstream processing or sharing with teammates.