Human in the Loop MCP

A VS Code extension that implements an MCP (Model Context Protocol) server for human-in-the-loop interactions. It allows any MCP-compatible AI agent to request user input, confirmations, or selections directly within VS Code.

What is Human in the Loop?

Human in the Loop (HITL) is a pattern where AI agents can pause their execution to request human input, verification, or decision-making. This is essential for:

• 🔐 Security: Getting explicit permission before destructive operations
• 🎯 Accuracy: Clarifying ambiguous requirements with the user
• 🔧 Flexibility: Collecting dynamic input that can't be predicted
• ✅ Trust: Keeping humans informed and in control of AI actions

Features

Installation

From VS Code Marketplace

1. Open VS Code
2. Go to Extensions (Ctrl+Shift+X / Cmd+Shift+X)
3. Search for "Human in the Loop MCP"
4. Click Install

From Source

git clone https://github.com/DercasDrol/human-in-the-loop-mcp.git
cd human-in-the-loop-mcp
npm install
npm run compile
# Press F5 in VS Code to launch debug mode

Quick Start

1. Configure Your AI Agent

Add the MCP server to your project's .vscode/mcp.json:

{
  "servers": {
    "human-in-the-loop": {
      "url": "http://127.0.0.1:3847/mcp"
    }
  }
}

Note: The extension automatically detects this configuration and starts the server on the specified port.

2. Use the Sidebar Panel

1. Click the Human in the Loop icon in the Activity Bar
2. When your AI agent sends a request, it appears in the panel
3. A countdown timer shows the remaining time to respond
4. Enter your response or click the appropriate button

Keyboard Shortcuts

Tip: Use Shift+Enter to write multi-line responses, then Enter to send.

Available MCP Tools

Your AI agent can use these tools to interact with users:

ask_user_text

Request free-form text input from the user.

Use for: API keys, file paths, custom names, descriptions, clarifications

{
  "name": "ask_user_text",
  "arguments": {
    "title": "API Key Required",
    "prompt": "Please enter your **OpenAI API key**:\n\nYou can find it at [platform.openai.com](https://platform.openai.com/api-keys)",
    "placeholder": "sk-..."
  }
}

ask_user_confirm

Request a Yes/No confirmation from the user.

Use for: Destructive operations, permission requests, verification

{
  "name": "ask_user_confirm",
  "arguments": {
    "title": "Confirm Deletion",
    "message": "Are you sure you want to delete the following files?\n\n- `src/old-module.ts`\n- `tests/old-module.test.ts`\n\n⚠️ **This action cannot be undone.**"
  }
}

ask_user_buttons

Present multiple options for the user to choose from.

Use for: Language selection, action menus, configuration choices

{
  "name": "ask_user_buttons",
  "arguments": {
    "title": "Select Framework",
    "message": "Which framework would you like to use for this project?",
    "options": [
      { "label": "React", "value": "react" },
      { "label": "Vue.js", "value": "vue" },
      { "label": "Angular", "value": "angular" },
      { "label": "Svelte", "value": "svelte" }
    ]
  }
}

Settings

Configure the extension in VS Code Settings (Ctrl+,):

Sound Types

• default - Standard notification (short)
• chime - Soft chime (short)
• ping - Quick ping (short)
• bell - Bell with harmonics (medium)
• notification - Two-tone ascending (medium)
• alert - Attention-grabbing beeps (long)
• melody - Musical phrase C-E-G-C (long)

Architecture

┌─────────────────┐     ┌──────────────────┐
│    AI Agent     │────▶│   MCP Server     │
│ (Any MCP Client)│◀────│  (HTTP/JSON-RPC) │
└─────────────────┘     └────────┬─────────┘
                                 │
                                 ▼
                        ┌──────────────────┐
                        │   VS Code        │
                        │   Extension      │
                        └────────┬─────────┘
                                 │
                                 ▼
                        ┌──────────────────┐
                        │  Sidebar Panel   │
                        │  (User Interface)│
                        └──────────────────┘

Session Isolation

Each VS Code window:

• Runs its own HTTP server on a unique port
• Has its own sidebar panel
• Handles requests independently

This enables working on multiple projects simultaneously without conflicts.

Compatibility

This extension works with any AI agent or tool that supports the Model Context Protocol (MCP), including:

• VS Code extensions with MCP support
• CLI tools implementing MCP client
• Custom AI agents using MCP SDK
• Any HTTP client that can send JSON-RPC requests

Commands

Privacy & Security

This extension:

• ✅ Runs entirely locally on your machine
• ✅ Does not collect any telemetry or analytics
• ✅ Does not send data to external servers
• ✅ All communication is between VS Code and your local AI agent
• ✅ By default binds to 0.0.0.0 for WSL/Remote compatibility (configurable to 127.0.0.1)

WSL & Remote Development

This extension fully supports VS Code Remote Development:

• WSL (Windows Subsystem for Linux)
• Remote - SSH
• Dev Containers
• GitHub Codespaces

How it works

When working in a remote environment, the extension runs in the workspace context (remote machine). VS Code automatically forwards the MCP server port to your local machine.

Bind Address Setting

If you're only working locally and want stricter network isolation, you can change humanInTheLoop.bindAddress to 127.0.0.1.

Requirements

• VS Code 1.100.0 or higher
• Node.js 20+ (for development; VS Code includes its own runtime)

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

1. Fork the repository
2. Create your feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

Release Process

This project uses GitHub Actions for automated releases:

1. Update version in package.json
2. Create and push a version tag: git tag v1.0.1 && git push --tags
3. GitHub Actions will automatically:
   • Build the extension
   • Create a GitHub Release with the VSIX
   • Publish to VS Code Marketplace (if VSCE_PAT secret is configured)
   • Publish to Open VSX Registry (if OVSX_PAT secret is configured)

License

MIT License - see LICENSE for details.

Changelog

See CHANGELOG.md for version history.

Made with ❤️ for the AI community