Undetectable AI MCP Server

Humanize text and detect AI writing directly from Claude, Cursor, ChatGPT, or any MCP-compatible assistant — connected to your Undetectable AI account in one click.

Server URL

https://undetectable.ai/api/mcpStreamable HTTP

What is this?

The Model Context Protocol (MCP) lets AI assistants call external tools. Our MCP server exposes Undetectable AI's humanizer and AI detector as tools your assistant can use mid-conversation: draft with your assistant, humanize with your account's credits, and check any text against the detectors — without leaving the chat. Connect by signing in with your Undetectable AI account (recommended) or with your API key.

Connect your client

Which app are you connecting?

Pick your MCP client for exact setup steps.

Sign-in vs. API key — which one?

	Sign in with Undetectable AI	API key
Best for	Personal use in Claude, Cursor, ChatGPT — one-click connect.	Headless setups, CI, shared machines.
Setup	Paste the server URL, sign in in your browser.	Paste the server URL and your API key.
Credentials	Short-lived tokens managed automatically.	A static key you copy from undetectable.ai/develop.
Revoking access	Remove the connector in your MCP client; its refresh token lapses 30 days after last use. There is no self-serve revocation page yet — contact support to cut off a connection immediately.	Regenerate the key on undetectable.ai/develop.

Available tools

humanize_text

Rewrites AI-generated text so it reads naturally and passes AI detectors. Costs one credit per input word, billed to the connected account; the result is also saved to your document history.

Input

Parameter	Type	Description
content	string, required	The AI-generated text to humanize (50-30,000 characters). Costs one credit per word.
readability	string, optional	Target reading level: "High School", "University" (default), "Doctorate", "Journalist" or "Marketing".
purpose	string, optional	What the text is for: "General Writing" (default), "Essay", "Article", "Marketing Material", "Story", "Cover Letter", "Report", "Business Material" or "Legal Material".
strength	string, optional	How aggressively to rewrite: "Quality", "Balanced" (default) or "More Human".
model	string, optional	Humanizer engine: "v2" (default), "v10", or the premium "v11" / "v11sr" engines on paid plans.

Output

Field	Type	Description
humanized_text	string	The rewritten, humanized text.
document_id	string	Saved document id — the result also appears in your Undetectable AI document history.
model_used	string	Engine that actually ran (premium requests may fall back).
model_downgraded	boolean, optional	Present (true) only when a premium engine was requested but the default engine ran instead.
input_word_count	number	Words in the input = credits the call cost.

detect_ai_text

Analyzes text with Undetectable AI's multi-engine detector and returns the likelihood it was AI-generated. Free — it never consumes credits.

Input

Parameter	Type	Description
text	string, required	The text to analyze for AI authorship (50-30,000 characters). Free.

Output

Field	Type	Description
ai_percentage	number	Probability (0-100) the text is AI-generated.
human_percentage	number	Probability (0-100) the text is human-written.
verdict	string	The detector's own verdict label.
detector_scores	object	Per-engine AI scores (GPTZero, ZeroGPT, Copyleaks, and more).
detection_id	string	Id of this detection run, for reference and support.

Try it

Once connected, just ask your assistant — it picks the right tool automatically:

text

Humanize this paragraph so it passes AI detection, then run it through the detector to confirm.

Need help?

If your client can't connect or a tool call fails, reach out and include the client you're using and the error you see.

Contact support