Add a human review checkpoint to an AI coding workflow in Cursor or Cline so the AI asks you before making assumptions.
Use the browser-based interface to get human feedback prompts over SSH or inside a WSL environment without a desktop app.
Save frequently used feedback responses as preset prompts so you can quickly guide the AI with one click.
Attach a screenshot or clipboard image to your feedback so the AI can see visual context when adjusting its approach.
Add a JSON block to your AI tool's MCP settings and run the server via uvx, no repo clone needed.
MCP Feedback Enhanced is a tool that adds a human feedback step into AI-assisted coding workflows. When an AI coding assistant is working on a task, it normally makes a series of decisions on its own before showing you the result. This tool interrupts that process at defined points to ask you a question or show you a summary, collect your response, and then pass that feedback back to the AI so it can adjust what it does next. The goal is to reduce the number of incorrect assumptions the AI makes by checking in with you rather than guessing. The tool works as an MCP server, which is a plugin format used by several AI coding environments including Cursor, Cline, Windsurf, and others. Once installed and configured, it intercepts AI calls and opens an interface where you can type a response, select from preset prompts you have saved, or attach an image. Your response is sent back to the AI in real time over a WebSocket connection. There are two interface options. The first is a desktop application built on a framework called Tauri, which runs natively on Windows, macOS, and Linux. The second is a browser-based web interface that works without any desktop GUI, which makes it suitable for remote development over SSH or for Windows Subsystem for Linux (WSL) environments. The tool detects the environment automatically and chooses the appropriate interface, or you can set it explicitly through configuration. Both interfaces offer the same features: saving frequently used prompts, setting a timer to auto-submit a response after a set number of seconds, session history that you can export in several formats, image uploads by drag-and-drop or clipboard paste, audio notifications, and multi-language support in English, Traditional Chinese, and Simplified Chinese. Installation is done by adding a short JSON configuration block to your AI tool's MCP settings. The server itself is installed and run via a Python packaging tool called uvx, so you do not need to clone this repository to use it. The repository is an enhanced fork of an earlier project by Fabio Ferreira called interactive-feedback-mcp.
← minidoracat on gitmyhub — every repo by this author, as a profile.
Verify against the repo before relying on details.