user-feedback

This MCP server provides a human-in-the-loop workflow for collecting user feedback during tool-assisted interactions (e.g., with Cline or Cursor). The server exposes a user_feedback tool that prompts a user for input before completing a task, enabling more nuanced testing and validation of complex interactions. When integrated, the MCP prompts appear in the web UI and can be routed through the host application so that feedback is gathered prior to task completion. To try it out, start the server using uv and open the web interface (default port is 5173). The available tool call shows a project_directory and a summary payload that will be sent to the user for feedback.

Once running, you can invoke the user_feedback tool with a request payload such as project_directory: your project path, and a summary describing the requested changes. The server will present the UI to collect user input and relay it back to the host application for processing.

Prerequisites:

• Python installed (for uv) or ensure uv is available as specified below.
• Internet access to install uv (if not already installed).

Installation steps:

1. Install uv globally (as shown in the README):
   • Windows: pip install uv
   • Linux/macOS: curl -LsSf https://astral.sh/uv/install.sh | sh
2. Clone the MCP server repository locally: git clone https://github.com/mrexodia/user-feedback-mcp.git cd user-feedback-mcp
3. Run the MCP server locally via uv to start the development server: uv run fastmcp dev server.py
4. Open the web interface in your browser at http://localhost:5173 and test the user_feedback tool.

Optional Cline integration (example):

• In Cline, add the MCP server configuration as shown in the README, pointing to the uv command with the appropriate directory and server.py entry point, and set a suitable timeout and autoApprove criteria.

Tips and notes:

• The example .user-feedback.json config shows a command and an execute_automatically flag. You can adapt this to automatically run a multi-step command if needed.
• If you integrate with Cline, ensure the directory path and server.py entry point are correct for your environment.
• The autoApprove setting can streamline testing by automatically approving the user_feedback tool when the specified tool name is triggered.
• If you encounter timeouts, adjust the timeout value in the MCP server configuration to accommodate longer interactions.
• The web UI is typically available at http://localhost:5173 when running locally; ensure no other process is occupying that port.
• This server is designed to work with human-in-the-loop workflows in tools like Cline and Cursor for testing complex user interactions.