API & MCP · Standard and Pro

humanize.fast inside the AI you already use

One URL, one token, three tools: humanize, detect, find_symbols. Runs started from an AI count against your monthly words exactly like runs started on the site, with the same quality guards.

Setup, under a minute

Create your token under Settings → API & MCP (Standard or Pro plan). It is shown once.
Add the server to your client with the snippets below.
Ask: "Humanize this paragraph at the deep level", "Is this text AI-written?", "Are there hidden characters in this?"

Claude Desktop · Cursor · VS Code · Windsurf (mcp.json)

{
  "mcpServers": {
    "humanize.fast": {
      "url": "https://humanize.fast/mcp",
      "headers": { "Authorization": "Bearer hf_…your token…" }
    }
  }
}

Claude Code

claude mcp add --transport http humanize.fast https://humanize.fast/mcp \
  --header "Authorization: Bearer hf_…your token…"

Codex (config.toml)

[mcp_servers."humanize.fast"]
url = "https://humanize.fast/mcp"
http_headers = { Authorization = "Bearer hf_…your token…" }

Tools

humanize text, level (normal|deep; quick, complete and technical still accepted), custom_instructions?

Rewrite AI-generated text so it reads as natural human writing. Charges credits like the website; returns the rewrite and the detector score before and after.

detect text

Score a text for AI authorship with Pallas, our own detector, in 39 languages. Free within the account's allowance.

find_symbols text

Find zero-width characters, homoglyphs and other invisible Unicode in a text and get a cleaned copy. Free.

REST API

The same tools over plain HTTP, same token.

curl https://humanize.fast/api/v1/tools/humanize \
  -H "Authorization: Bearer hf_…your token…" \
  -H "Content-Type: application/json" \
  -d '{"text": "Your AI-written paragraph…", "level": "deep"}'

GET  https://humanize.fast/api/v1/me      # plan, credits remaining
GET  https://humanize.fast/api/v1/tools   # the tool list
POST https://humanize.fast/api/v1/tools/detect        {"text": "…"}
POST https://humanize.fast/api/v1/tools/find_symbols  {"text": "…"}

Responses are the same JSON the website receives: for humanize, humanized_text, ai_score_before, ai_score_after, credits_used; errors carry an error line and, where relevant, a code. Rate limit: 60 requests a minute. Word limits per rewrite follow your plan.

What the scores mean

Scores come from Pallas, our own detector, measured in 39 languages. They are our model's opinion, not the verdict of GPTZero, Turnitin or any other service, and we do not guarantee any result on those.