Connect TruthScan AI Image Detection to Claude, Cursor, and any MCP client
Run AI-image detection as a tool directly inside your AI assistants. No glue code, no custom integration — point your MCP client at TruthScan and start detecting AI-generated images in conversation.
https://truthscan.com/api/mcpHow it works
The Model Context Protocol (MCP) lets AI assistants call external tools through a standard interface. TruthScan exposes its AI-image detector as an MCP tool, so any MCP-compatible client — Claude Desktop, Cursor, ChatGPT, and others — can submit an image and receive an AI-likelihood score without writing integration code. Connect in one of two ways: a static API key for fast, headless setups, or OAuth 2.1 for per-user sign-in.
Set up your connection
What are you connecting with?
Tell us what you're connecting with and we'll walk you through it step by step. All clients reach the same MCP server over streamable HTTP.
Which mode should I use?
Both modes reach the same tools. Pick based on who is running the client and how tightly you need to control access.
| Consideration | API key | OAuth 2.1 |
|---|---|---|
| Best for | Fast setup, headless / CI environments | Multi-user tools with per-user consent |
| Who signs in | Single shared account | Each user signs in individually |
| Credentials in config | Bearer token pasted in headers | None — obtained interactively |
| Token lifetime | Long-lived until rotated | Auto-expiring, revocable per user |
| Setup effort | Lowest — paste and go | Browser sign-in on first connect |
Tools
TruthScan exposes its detectors as MCP tools. Your assistant can call them directly during a conversation.
detect_imageSubmits an image to TruthScan's detector and returns the AI-likelihood score. A companion tool, create_image_upload_url, returns a short-lived upload URL plus a file_ref so local files can be detected without hosting them publicly — PUT the bytes to that URL (and to storage_upload_url when it is returned, so the image is kept on the dashboard record), then pass the file_ref here.
Input
| Parameter | Type | Description |
|---|---|---|
| image_url | string (one of) | Publicly reachable URL of the image to analyze. |
| file_ref | string (one of) | Reference from create_image_upload_url, once the image has been uploaded. Provide exactly one of image_url or file_ref. |
Output
| Field | Type | Description |
|---|---|---|
| score | number (0–100) | AI-likelihood score. Higher means more likely AI-generated. |
| is_ai | boolean | Whether TruthScan classifies the image as AI-generated. |
| confidence | "Low" | "Medium" | "High" | Confidence level of the classification. |
| detection_result_id | string (optional) | Saved detection record id, viewable in your TruthScan dashboard history. |
| detector_id | string (optional) | Upstream detector job id for this analysis, for reference. |
Example result
{
"score": 92,
"is_ai": true,
"confidence": "High",
"detection_result_id": "det_9f2a1c7b",
"detector_id": "a1c7b9f2-..."
}More tools for video, audio, text, and PDF detection are coming soon.
Need help connecting?
Our team can help you wire TruthScan into your MCP client and pick the right authentication mode for your workflow.
Contact supportFrequently asked questions
Common questions about connecting to the TruthScan MCP server.