# GPTZero MCP for AI Agents AI Agent Connect > GPTZero lets you detect AI-generated text with confidence scores and pinpoint exactly which passages were likely written by a language model. It provides deep technical metrics like perplexity and burstiness to help you understand text variation. Instead of manual copy-pasting into web detectors, you can run these checks directly through your AI client to audit content authenticity in real-time. ## Overview - **Category:** artificial-intelligence - **Price:** Free - **Endpoint:** https://edge.vinkius.com/vk_preview_Y0bXy1KzXoResEbnjX7cvEDgsbjGgKLkQUAoqZai/ai-agent-connect - **Tags:** ai-detection, content-authenticity, academic-integrity, text-analysis, perplexity-scoring, machine-learning ## Description You're tired of copy-pasting paragraphs into different web detectors just to see if a student or a freelancer actually did the work. This Connector changes that by letting your AI agent handle the heavy lifting of content auditing directly. When you connect GPTZero, you can ask your agent to scan a whole document and get a breakdown of what's human and what's machine-made. It doesn't just give you a yes or no answer; it provides the underlying data like perplexity and burstiness scores so you can see the structural randomness of the writing. If a result seems off, you can send feedback back through the agent to sharpen the detection accuracy over time. It even pulls in the official interpretation guide so you can explain the why behind a score to stakeholders. It's a way to turn your AI client into a dedicated academic integrity or content authenticity auditor without ever leaving your workspace. You'll find this is a huge time-saver when managing large volumes of content or checking submissions, and it's one of the many high-quality integrations you can find in the Vinkius catalog. ## Tools ### get_current_user See the profile of the authenticated user. Use this to confirm which account is currently active. ### get_api_quotas Check how many character credits you have left. Use this to monitor your spending and limits. ### detect_ai_in_text Send a string to GPTZero to check for AI generation. This is the primary tool for auditing content. ### submit_prediction_feedback Send feedback on a specific detection result. This helps improve the accuracy of future scans. ## Prompt Examples **Prompt:** ``` Check this blog post for AI content: [insert text] ``` **Response:** ``` I've analyzed the text. **GPTZero** shows a 92% probability of AI generation. * **Perplexity:** 8.4 (Low) * **Burstiness:** Low This suggests the sentence structures are very predictable, which is common in machine-made content. It looks like it might have been generated by a model like Claude or Gemini. Would you like me to check another section? ``` **Prompt:** ``` How many credits do I have left? ``` **Response:** ``` Checking your account status... You currently have **4,200 character credits** remaining for this billing cycle. You're on the 'Standard' plan, and at your current rate, you should have enough for about 15 more full-page scans. ``` **Prompt:** ``` What does a perplexity score of 5 mean? ``` **Response:** ``` A perplexity score of 5 is very low. In the GPTZero system, this usually indicates that the text is highly predictable, which is a common trait of AI-generated content. **Quick Reference:** * **Score < 10:** Likely AI * **Score > 20:** Likely Human Human writing typically has much higher variety in word choice and structure. ``` ## Capabilities ### Scan text for AI generation Analyze raw strings to see if a model wrote them. ### Get perplexity and burstiness scores See how predictable and varied the sentence structure is. ### Submit prediction feedback Tell the system if it got a detection right or wrong. ### Check remaining API credits See how many character limits you have left. ### Get interpretation guides Understand the technical metrics for stakeholders. ### List supported models See which specific models the detector can identify. ### Verify API health Check if the connection to the detection engine is live. ## Use Cases ### Verifying student essays An educator pastes a submission and asks the agent to check for AI use. ### Blog post auditing A marketing manager asks to scan a draft to ensure it has a human touch. ### Compliance checks A legal officer asks the agent to scan a contract for automated content. ### Batch content review A content lead asks to check ten different articles for AI signatures at once. ## Benefits - Skip the manual copy-pasting by running scans directly through your AI client using detect_ai_in_text. - Get deep technical data like perplexity and burstiness to prove why a text looks like AI. - Monitor your spending and limits in real-time with get_api_quotas so you never hit a wall. - Improve detection accuracy over time by sending feedback via submit_prediction_feedback. - Explain complex scores to stakeholders using the official guide from get_interpretation_guide. ## How It Works The bottom line is you get a direct line to professional-grade AI detection without the manual browser round-trips. 1. Grab your API key from the GPTZero Dashboard under the API section. 2. Connect this Connector to your AI client using the Vinkius setup. 3. Ask your agent to scan any text for AI involvement and get the scores back. ## Frequently Asked Questions **Can GPTZero help me find out if a student used AI?** Yes, it provides a probability score and technical metrics that help you determine if a submission was likely generated by a language model. **How does GPTZero detect AI-generated text?** It analyzes factors like perplexity and burstiness to see how much variation and randomness are in the writing. **Can I use GPTZero to check my own marketing content?** Absolutely. It's a great way to ensure your brand's voice stays human and doesn't get lost in generic AI output. **Does GPTZero tell me which specific AI model was used?** It provides a probability score for AI involvement and highlights the specific passages that look most like machine output. **How do I know if a low perplexity score is a problem?** A low score means the text is very predictable. This is a common signal for AI, but you can use the interpretation guide to see how to explain this to others. **Is GPTZero good for large volumes of text?** Yes, because you can use your agent to scan text directly without having to manually copy and paste every single paragraph. **How do I find my GPTZero API Key?** Log in to your GPTZero dashboard, navigate to the **API** section, and copy your unique identifier. **What do the perplexity and burstiness scores indicate?** Perplexity measures randomness; low scores suggest predictability (likely AI). Burstiness measures structural variation; low variation is typical in AI-generated content. **Can I provide feedback if a result is incorrect?** Yes! Use the `submit_prediction_feedback` tool to provide the correct classification, which helps improve GPTZero's detection model.