Editorial review
Ask an assistant to screen an article draft, summarize detector signals, and identify passages that need human review.
MCP integration
AI Detector MCP for Claude, ChatGPT, and Codex
Analyze text inside your assistant through a secure Streamable HTTP connection with automatic OAuth for hosted connectors.
Hosted connectors need only the MCP URL and your Detecting AI sign-in. The install prompt remains available for clients that read a local MCP config and use an API key.
Hosted connector details
MCP URL
https://api.detecting-ai.com/api/mcp
Hosted authentication
OAuth 2.0 — sign in and approve in your browser
Tools
detect_ai_text, humanize_textWhat an AI Detector MCP server does
An AI Detector MCP server gives a compatible assistant permission to call a dedicated detection tool. The assistant sends only the text selected for analysis, receives structured detector output, and can explain the result in the context of the current task.
- Stay in the active conversationCheck a draft, submission, or document without moving text between separate browser tools.
- Use structured detector outputThe tool returns the detector version, processed word count, and detailed analysis for the assistant to interpret.
- Keep one subscription balanceMCP calls use the same detector word allowance as the Detecting AI REST API.
- Control when analysis runsThe assistant calls the detector only when the user requests analysis or approves a workflow that includes it.
From prompt to detector result
When a user asks whether text contains AI-writing signals, the client selects the Detecting AI tool and sends the supplied passage. The MCP server validates the connected account or local API key and detector quota, runs the selected detector version, records usage, and returns structured output to the client.
Available MCP tool
detect_ai_text: Analyzes text for AI-writing signals and returns detailed detector output.
| Input | Type | Status | Description |
|---|---|---|---|
text | string | Required | The passage or document section to analyze. |
version | v1 | v2 | v3 | Optional | Detector version. The default is v3. |
Compatible MCP clients
Hosted connectors discover OAuth automatically. Local clients can use the same Streamable HTTP endpoint with an API key in the Authorization bearer header.
Claude CodeAdd the remote HTTP server from the command line.
CodexReference the endpoint and API key environment variable in config.toml.
ClaudeAdd a custom connector and sign in to Detecting AI with OAuth.
ChatGPTEnable Developer Mode, add the MCP URL, and sign in with OAuth.
Connect from Claude or ChatGPT
Add Detecting AI as a custom connector and paste the public MCP URL. The client will open Detecting AI so you can sign in, review the requested permission and word-usage notice, and approve or decline the connection. You never need to copy an API key into a hosted connector.
- Open the client's connector settings. ChatGPT users must enable Developer Mode before adding a custom connector.
- Paste
https://api.detecting-ai.com/api/mcpas the connector URL. - Sign in to Detecting AI and approve the connection.
Public MCP endpoint
https://api.detecting-ai.com/api/mcpDo not append an API key, OAuth token, or any other credential to this URL.
Where AI Detector MCP fits
MCP works best when the assistant needs a focused external tool as part of a user-controlled workflow.
Education workflows
Review submitted writing inside an approved academic workflow while treating detector output as evidence for further review, not proof.
Content operations
Add detector checks to agent-assisted moderation, intake, or quality assurance processes without creating a separate integration layer.
Connect the detector to your assistant
For Claude Code and other local-config clients, create a Detecting AI API key, store it outside the project, and send it as a bearer token.
Public MCP endpoint
https://api.detecting-ai.com/api/mcpClaude Code setup
claude mcp add --transport http --scope user \
--header "Authorization: Bearer YOUR_API_KEY" \
detecting-ai https://api.detecting-ai.com/api/mcpAI Detector MCP questions
What is an AI Detector MCP server?
It is a remote Model Context Protocol server that lets compatible AI clients call a dedicated text detection tool with structured inputs and outputs.
Does it work with Claude, ChatGPT, and Codex?
Yes. Hosted Claude and ChatGPT connectors use OAuth, while Claude Code, Codex, and other local-config clients can continue using an API key.
What does the detector tool return?
It returns detailed detector analysis, the selected detector version, and the number of processed words.
Is an AI detection result proof of authorship?
No. Detection scores are probabilistic signals. They should be combined with context, policy, and human review before decisions are made.
How is MCP usage billed?
Successful detector calls use words from the same Detecting AI API subscription and detector balance used by REST requests.
Use AI Detector from your AI client
Paste the public MCP URL into a hosted connector and sign in, or use an API key from a local MCP client.