🚩 Flag MCP

✨ What is "Flag"?

In visual novels and Galgames, a "flag" (フラグ) is the moment a choice triggers a new story branch. One decision changes everything.

Flag MCP brings that same power to AI-assisted coding. When your AI assistant hits a crossroads, it doesn't guess — it raises a flag and waits for you to choose the route.

🎮 You're the protagonist. The AI waits at every branching point.

💎 Every flag shapes the route. No more speculative rewrites.

🚀 Rich interaction. Text, screenshots, annotations — your full arsenal.

This transforms AI coding from "hope it works" into a narrative where you hold the controller.

Scope of Application:

• Coding plans billed on a per-request basis.
• Developers who wish to control AI behavior.

🔥 Before & After

🎯 Core Features

• 🖥️ Dark Themed UI — A sleek native desktop dialog that fits your workflow
• ✅ Route Choices — Structured predefined options (checkbox-style)
• 💬 Free Text — When the predefined routes aren't enough, write your own script
• 📷 Rich Media Arsenal
  • Paste images from clipboard
  • Select local files
  • Screenshot + Built-in Annotator (rectangle, circle, arrow, pen, text, crop)
• 🖼️ Prompt Images — AI can show you images (local paths, file://, http(s)://)
• 🔒 Security First — Remote images validated, size-limited, async loaded
• 🎨 macOS Optimized — Proper icon handling and visual polish

📦 Installation

Prerequisites

• Python >= 3.11
• uv (recommended) or pip

Quick Install

git clone https://github.com/pauoliva/interactive-feedback-mcp.git
cd interactive-feedback-mcp
uv sync

⚙️ Configuration

Add to your MCP client configuration:

Cursor (mcp.json) / Claude Desktop (claude_desktop_config.json):

{
  "mcpServers": {
    "interactive-feedback": {
      "command": "uv",
      "args": [
        "--directory",
        "/path/to/interactive-feedback-mcp",
        "run",
        "server.py"
      ],
      "timeout": 900000,
      "autoApprove": ["interactive_feedback"]
    }
  }
}

⚠️ Note: Timeout is in milliseconds for Cursor (900000 = 15 min). Some clients use seconds — adjust accordingly.

🚩 The interactive_feedback Tool

Arguments

Returns

• Text feedback from user
• Optional image attachments (as MCP image content blocks)

🧙 Pro Tips

Recommended Agent Rules

Add this to your AI assistant's custom instructions:

If requirements are unclear, call interactive_feedback before implementing.
Present predefined options whenever possible — give the user clear route choices.
Before finishing a task, call interactive_feedback once more for final confirmation.

Environment Variables

🛡️ Security & Reliability

• ✅ Remote images validated by content-type (image/*)
• ✅ Large payloads rejected via size cap
• ✅ Async fetch keeps UI responsive
• ✅ Local files read only when explicitly referenced

📄 License

MIT License — fork it, flag it, ship it.