Prompt Suggestion MCP Server

A Model Context Protocol (MCP) server that helps any LLM (ChatGPT, Claude, etc.) instantly rewrite disappointing responses into one clean, high-quality prompt.

This MCP server analyzes the recent conversation, identifies what the user actually wanted, and generates a single, first-person improved prompt tailored to produce a better response from the assistant.

No noise.
No multiple choices.
Just one perfect next prompt — every time.

✨ Features

🎯 Smart Failure Understanding

The server reads:

• The recent conversation window
• The last user message
• An optional user comment (why they disliked the answer)

And analyzes:

• What the user was trying to accomplish
• Why the assistant’s prior answer failed
• What information or structure was missing

✍️ High-Quality Prompt Rewriting

The MCP server generates a single optimized prompt that:

• Is written fully in first person
  • (“Explain to me…”, “Give me…”, “Help me…”)
• Is self-contained
  • The prompt works even if the model never saw the past conversation
• Is clear and specific
• Avoids generic references like “the user” or “previous discussion”
• Adds helpful structure when needed (steps, examples, analogies, UX guidelines, etc.)

Optional: Up to 2 lightweight alternatives may be included.

🧠 Consistent Output

No matter how the user triggers the tool —
“use the tool”, “suggest a better prompt”, “rewrite this”, etc. —
the server always returns one single best prompt, not multiple choices.

🔗 Connect the MCP Server to ChatGPT (Developer Mode)

1️⃣ Enable Developer Mode

Open ChatGPT:

Settings → Apps & Connectors → Advanced Settings → Enable Developer Mode

2️⃣ Add the MCP Server

Go to:

Apps & Connectors → Create → New MCP Connector

Fill in the following:

Check the trust box → Click Create

3️⃣ Refresh ChatGPT

Now click the + button next to the message box.
Under More, you will see the MCP tool with the name you assigned.

Your prompt-rewriting MCP server is now active 🎉

🧠 How to Use the MCP Tool

At any time, simply type:

• “Use the tool and give me a better prompt.”
• “Rewrite this using the MCP server.”
• “Fix my prompt using the tool.”

ChatGPT will:

1. Detect your request
2. Call your MCP server
3. Ask for permission
4. Return a clean, improved prompt

Every result contains:

• A summary of what went wrong
• Root-cause reasons
• One single improved prompt (the one you should send next)
• Optional small alternative suggestions
• A confidence score

🧩 Output Format

The MCP server always returns:

{
  "summary": "...",
  "root_causes": ["..."],
  "suggested_prompt": "THE BEST FIRST-PERSON PROMPT",
  "alternatives": ["..."],
  "confidence": "0.0 - 1.0"
}