← Discover MCPs and Agents
p
MCPAI & MLGitHub

pro-prompter

Rewrites a rough request into a FAANG-grade prompt and runs it - 15 specialised task types. Free MCP plugin for Claude and ChatGPT, zero extra credits.

Links

README

From the repo.

Pro Prompter 🧠

A free, open-source plugin for Claude & ChatGPT that rewrites your rough requests into FAANG-grade engineered prompts and runs them — using the AI you're already talking to, at zero extra credits.

One MCP server. Works in both Claude and ChatGPT.

What it does

InterfaceWhat it does
pro_prompt (tool)Engineer your request into a top-tier prompt and execute it — best answer in one step
refine_prompt (tool)Return just the polished prompt, to copy / save / reuse
recall_prompts · clear_memory (tools)Per-session memory of your requests
pro_prompt (prompt)A pickable slash-command version — Claude only

How "zero extra credits" actually works (the honest bit)

An MCP plugin can't secretly rewrite your message before the AI reads it, and it can't call an AI of its own (Claude/ChatGPT don't allow that in 2026). So Pro Prompter does something smarter:

Each tool returns a world-class meta-prompt (plain text — instant and free). The AI you're already chatting with reads it and does the rewriting + answering in the same turn. No API key, no second bill — it rides on your existing Claude/ChatGPT.

The trade-off: you invoke it explicitly (say "Pro Prompter: …" or pick the slash-command). It can't auto-upgrade every message — that's a hard platform rule, not a missing feature.

The rubric it injects is distilled from Anthropic's and OpenAI's official prompt-engineering guides: preserve your intent verbatim, add the "why", positive constraints, explicit output format, model-class-aware reasoning, a complexity gate (no over-engineering simple asks), and strict token economy (structure, not bloat).

Use it

In Claude or ChatGPT, once connected:

  • Refine + run: Pro Prompter: build a landing page for my bakery
  • Refine only: Pro Prompter, refine only: build a landing page for my bakery
  • Memory: Pro Prompter, what have I asked this session?

Deploy it (free)

Prereqs: Node.js 18+, a free Cloudflare account.

cd "path\to\pro-prompter"
npm install
npx wrangler deploy      # logs into Cloudflare, prints your live URL

Your connector URL is the printed address + /mcp:

https://pro-prompter.<your-account>.workers.dev/mcp

Open the base URL in a browser to confirm the "✔ live" page.

Put it on GitHub + auto-deploy (optional)

git init; git add .; git commit -m "Pro Prompter"
git branch -M main
git remote add origin https://github.com/<you>/pro-prompter.git
git push -u origin main

Then add a CLOUDFLARE_API_TOKEN repo secret (Settings → Secrets → Actions) and every push auto-deploys via .github/workflows/deploy.yml.


Connect it

Claude (any plan, incl. Free)

Settings → ConnectorsAdd custom connector → paste the /mcp URL → Add. Leave OAuth blank. In a chat, enable it via the + menu. Claude also shows the pro_prompt slash-command.

ChatGPT (Plus/Pro/Business+, desktop web)

Settings → ConnectorsAdvancedDeveloper mode → add the /mcp URL, Auth None. Enable it in a chat via the tools menu. (ChatGPT ignores the slash-command prompt; use the tool.)

⚠️ Use the URL ending in /mcp. The bare ...workers.dev is just a status page, not the MCP endpoint.


Extend it

The whole rubric lives in src/index.ts (RUBRIC, analyze, buildRefineAndRun). Tweak the rubric or add task types, then npx wrangler deploy (or git push).

License

MIT © 2026 Pranay Mahendrakar — free to use, modify, and share.

Collected info

  • 0 stars
  • Language: TypeScript
  • Source updated: 7/15/2026

Config for your environment

Replace {MCP_ENDPOINT_URL} with this MCP’s endpoint URL (from its repo or docs above). No API key — you connect directly.

Tool

OS

Config file: ~/.cursor/mcp.json

{
  "mcpServers": {
    "mcp-server": {
      "url": "{MCP_ENDPOINT_URL}"
    }
  }
}

Paste into mcpServers in the config file. Restart Cursor after saving.

If this MCP is also published on mcpchannel.ai, you can subscribe from Browse and use the gateway config there instead.