MCP Server + REST API

GPT Scrambler MCP Server — Humanize AI Text in Claude & Cursor

Install @gptscrambler/mcp-server and let your AI agent humanize drafts inline. Same quality as the website — built for developers, writers, and power users who live in Claude Desktop, Cursor, or Claude Code.

Open-source · stdio transport · Bearer API key · Works with Turnitin/GPTZero-grade output

Get API Key Setup in 2 minutes View Docs

npm package @gptscrambler/mcp-server — MIT licensed on GitHubnpm GitHub

What is the GPT Scrambler MCP Server?

The Model Context Protocol (MCP) is an open standard that lets AI assistants call external tools — search, databases, or in our case, a text humanization API. When you add GPT Scrambler as an MCP server, Claude or Cursor can invoke humanize_text on your behalf instead of you copying AI drafts into a browser.

Under the hood it is the same GPT Scrambler engine millions of users rely on for bypassing AI detectors: identical prompts, same quality tiers (standard and aggressive modes), and the same privacy guarantees. The MCP layer is a thin stdio wrapper around our public REST API — nothing proprietary or locked to one IDE.

Whether you connect via MCP in Claude Desktop or hit POST /api/v1/humanize from a Python script, you get the same humanized output. That makes GPT Scrambler a practical building block for content pipelines, academic writing workflows, and agentic coding setups.

How it works

Get your API key

Connect to your agent

Paste the JSON config into Claude Desktop, Cursor, or Claude Code — one npm package, no global install.

Humanize in context

Ask: "humanize this paragraph." The agent calls GPT Scrambler and returns polished text in the same thread.

Use cases

Humanize inside Claude or Cursor

Draft with your AI, then say “humanize this paragraph with GPT Scrambler.” No copy-paste to a browser tab — the agent calls the MCP tool and returns polished text in the same thread.

Batch content pipelines

Wire the REST API into a Node or Python script to humanize hundreds of product descriptions, email variants, or support macros overnight. Up to 3,000 words per request.

Editor & doc workflows

Connect in Cursor or Claude Code to refine README sections, blog drafts, or PR descriptions before you commit. Check balance with get_balance before large jobs.

Quality gate before publish

Run AI-generated copy through GPT Scrambler before it hits your CMS, newsletter, or LMS. Same engine as the web app — built to pass GPTZero, Turnitin, and Copyleaks.

MCP tools reference

After connecting, your agent sees two tools. Full parameter docs: MCP setup guide.

humanize_text

Send AI-generated text, get humanized output + words used

text (required), language?, mode? (standard | aggressive)

get_balance

Remaining words, plan name, and reset date

none

Example prompts for your AI

Once MCP is connected, paste any of these into Claude Desktop, Cursor chat, or Claude Code:

“Humanize this paragraph using GPT Scrambler: [paste text]”
“Make this ChatGPT draft sound more natural — use the humanize_text tool”
“Check my GPT Scrambler API balance, then humanize this email”
“Rewrite this in aggressive mode to bypass AI detection”

Connection methods

MCP docs API reference

Claude Desktop (Remote URL)

Remote MCP

Connect by URL — no npx, no local install. Paste the config below into your Claude Desktop config file, save, and restart Claude Desktop.

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json · Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "gpt-scrambler": {
      "url": "https://gptscrambler.com/api/mcp",
      "headers": {
        "Authorization": "Bearer sk_live_your_key_here"
      }
    }
  }
}

Cursor (Remote URL)

Remote MCP

Paste the config below into .cursor/mcp.json (project-level) or ~/.cursor/mcp.json (global). Cursor picks up changes automatically — no restart needed.

Project: .cursor/mcp.json · Global: ~/.cursor/mcp.json

{
  "mcpServers": {
    "gpt-scrambler": {
      "url": "https://gptscrambler.com/api/mcp",
      "headers": {
        "Authorization": "Bearer sk_live_your_key_here"
      }
    }
  }
}

Claude Code

MCP

Run the command below in your terminal. After it completes, restart Claude Code for the tools to appear.

Run in terminal (then restart Claude Code):

claude mcp add --transport http gpt-scrambler https://gptscrambler.com/api/mcp \
  --header "Authorization: Bearer sk_live_your_key_here"

Claude Desktop (local npx)

MCP (stdio)

Prefer a local process? Paste the config into your Claude Desktop config file and restart Claude Desktop. Requires Node.js 18+.

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json · Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "gpt-scrambler": {
      "command": "npx",
      "args": ["-y", "@gptscrambler/mcp-server"],
      "env": {
        "GPTSCRAMBLER_API_KEY": "sk_live_your_key_here"
      }
    }
  }
}

Cursor (local npx)

MCP (stdio)

Local npx method for Cursor. Paste the config into .cursor/mcp.json. Requires Node.js 18+.

Project: .cursor/mcp.json · Global: ~/.cursor/mcp.json

{
  "mcpServers": {
    "gpt-scrambler": {
      "command": "npx",
      "args": ["-y", "@gptscrambler/mcp-server"],
      "env": {
        "GPTSCRAMBLER_API_KEY": "sk_live_your_key_here"
      }
    }
  }
}

REST API (curl)

REST

Call the API from Python, Node, Zapier, or any HTTP client. Same API key as MCP.

curl -X POST https://gptscrambler.com/api/v1/humanize \
  -H "Authorization: Bearer sk_live_your_key_here" \
  -H "Content-Type: application/json" \
  -d '{"text": "Your AI-generated text here...", "mode": "standard"}'

ChatGPT developer mode / Actions: remote MCP over HTTP is on our roadmap. Today use Claude Desktop, Cursor, Claude Code, or the REST API from any language.

Why use the API / MCP server?

Prompts stay in a black box — no competitor can copy your results

Same quality as the website — identical model and prompts

No context switching — stay in Claude, Cursor, or your code

100,000 words/month — enough for serious workflows

Simple Bearer token auth — one key, any language

200-word API/MCP trial separate from web; paid API plan uses your main balance

Open-source MCP server you can audit on GitHub

Works with standard and aggressive humanization modes

API Pricing

API Plan

FAQ

Which npm package should I use?

Install @gptscrambler/mcp-server via npx (see connection examples below). The older unscoped package gptscrambler-mcp-server is deprecated.

Who is this for?

Developers building AI pipelines, power users working in Claude Desktop or Cursor, and anyone who wants to humanize text without switching to a browser.

How is this different from the website?

It's the same engine, same prompts, same quality — delivered through an API or MCP protocol so you can use it directly inside your AI agent or code.

Does it work with ChatGPT?

ChatGPT Actions and remote MCP connectors are planned for Phase 2. Today it works with Claude Desktop, Claude Code, Cursor, and any REST API client.

Does Claude Code support this MCP server?

Yes. Add the same @gptscrambler/mcp-server config to ~/.claude.json or a project .mcp.json file. Claude Code can then call humanize_text from your terminal workflow.

What happens when I run out of words?

The API returns a 402 error with your remaining balance and a link to top up or upgrade your plan.

Is my text kept private?

Yes. Your input text is not logged or stored permanently — same privacy policy as the web interface.

Ready to integrate?

Generate your free trial key and humanize your first text in under 5 minutes. Read the step-by-step MCP guide on our blog.

Get API Key MCP Docs API Reference