Organize Interactive Human-in-the-Loop Workflows
User Feedback MCP adds a human-in-the-loop step to Cline or Cursor, letting you review AI changes before a task is marked done.
Why it matters
Streamline the development of desktop applications by organizing complex, interactive human-in-the-loop testing workflows. This asset facilitates seamless feedback integration for enhanced testing and development cycles.
Outcomes
What it gets done
Configure interactive human-in-the-loop workflows for tools like Cline and Cursor.
Request and manage user feedback during the development process.
Automate command execution and project-specific configurations.
Integrate multi-step commands and task execution with auto-approval capabilities.
Install
Add it to your toolbox
Run in your project directory:
curl -fsSL https://spark.entire.vc/get/vb-user-feedback | bash Capabilities
Tools your agent gets
Requests feedback from users with project directory and summary parameters
Overview
User Feedback MCP Server
User Feedback MCP gives AI coding assistants like Cline or Cursor a way to pause and ask the user for feedback via a UI before marking a task complete, instead of assuming the work is done. Use it when a task needs manual verification, such as desktop UI testing, before an assistant should consider it finished; not suited to fully unattended, no-human workflows.
What it does
User Feedback MCP is a simple MCP server that enables a human-in-the-loop workflow in tools like Cline and Cursor. Instead of an AI assistant marking a task complete on its own judgment, it can call the user_feedback tool to pause and ask the user for feedback first - particularly useful when developing desktop applications that require complex, hands-on testing an AI cannot fully verify itself.
When to use - and when NOT to
Use this when you want an extra checkpoint before an AI coding assistant declares a task finished, especially for changes that need manual testing or visual confirmation, like desktop UI behavior. A custom prompt instruction ("Before completing the task, use the user_feedback MCP tool to ask the user for feedback") ensures the assistant actually calls it. It is not useful for fully autonomous, unattended workflows where no human is available to respond to the feedback prompt, since the assistant will wait on that input.
Capabilities
user_feedback: takes aproject_directoryand asummaryof what was done, and opens a feedback UI for the user to review and respond before the assistant proceeds.- Saving configuration through the feedback UI creates a
.user-feedback.jsonfile with acommand(e.g.npm run dev) and anexecute_automaticallyflag; when enabled, the command runs automatically on startup without a manual click.
How to install
Install uv globally (pip install uv on Windows, or curl -LsSf https://astral.sh/uv/install.sh | sh on Linux/Mac), then clone this repository. In Cline, open MCP Servers -> Installed -> Configure MCP Servers, and add:
{
"mcpServers": {
"github.com/mrexodia/user-feedback-mcp": {
"command": "uv",
"args": ["--directory", "c:\\MCP\\user-feedback-mcp", "run", "server.py"],
"timeout": 600,
"autoApprove": ["user_feedback"]
}
}
}
For development and testing, run uv run fastmcp dev server.py, which opens a web interface at http://localhost:5173 to interact with the tool directly.
Who it's for
Developers using AI coding assistants like Cline or Cursor who want a deliberate checkpoint for human review - especially when building desktop apps or features that need manual testing before the assistant considers the work done.
Source README
User Feedback MCP
Simple MCP Server to enable a human-in-the-loop workflow in tools like Cline and Cursor. This is especially useful for developing desktop applications that require complex user interactions to test.
Prompt Engineering
For the best results, add the following to your custom prompt:
Before completing the task, use the user_feedback MCP tool to ask the user for feedback.
This will ensure Cline uses this MCP server to request user feedback before marking the task as completed.
.user-feedback.json
Hitting Save Configuration creates a .user-feedback.json file in your project directory that looks like this:
{
"command": "npm run dev",
"execute_automatically": false
}
This configuration will be loaded on startup and if execute_automatically is enabled your command will be instantly executed (you will not have to click Run manually). For multi-step commands you should use something like Task.
Installation (Cline)
To install the MCP server in Cline, follow these steps (see screenshot):
- Install uv globally:
- Windows:
pip install uv - Linux/Mac:
curl -LsSf https://astral.sh/uv/install.sh | sh
- Windows:
- Clone this repository, for this example
C:\MCP\user-feedback-mcp. - Navigate to the Cline MCP Servers configuration (see screenshot).
- Click on the Installed tab.
- Click on Configure MCP Servers, which will open
cline_mcp_settings.json. - Add the
user-feedback-mcpserver:
{
"mcpServers": {
"github.com/mrexodia/user-feedback-mcp": {
"command": "uv",
"args": [
"--directory",
"c:\\MCP\\user-feedback-mcp",
"run",
"server.py"
],
"timeout": 600,
"autoApprove": [
"user_feedback"
]
}
}
}
Development
uv run fastmcp dev server.py
This will open a web interface at http://localhost:5173 and allow you to interact with the MCP tools for testing.
Available tools
<use_mcp_tool>
<server_name>github.com/mrexodia/user-feedback-mcp</server_name>
<tool_name>user_feedback</tool_name>
<arguments>
{
"project_directory": "C:/MCP/user-feedback-mcp",
"summary": "I've implemented the changes you requested."
}
</arguments>
</use_mcp_tool>
FAQ
Common questions
Discussion
Questions & comments · 0
Sign In Sign in to leave a comment.