# GPTZero MCP MCP > GPTZero analyzes text for AI generation, providing deep technical metrics like perplexity and burstiness scores with high confidence. This MCP detects if content came from a language model or was written by a human, highlighting exactly which passages might be synthetic. ## Overview - **Category:** artificial-intelligence - **Price:** Free - **Tags:** ai-detection, content-authenticity, academic-integrity, text-analysis, perplexity-scoring, machine-learning ## Description Content authenticity used to be a painful manual process. You'd copy chunks of text into various web detectors, hoping they gave you clear answers—and usually getting vague warnings instead. This MCP changes that. Your agent takes the raw text and runs it through advanced detection protocols, giving you hard data on its origin. It doesn't just say 'AI generated'; it tells you *how* likely it is by calculating structural metrics like perplexity and burstiness. Through this single connection, your AI client acts as a dedicated content auditor for everything from student papers to marketing copy. Need to check quotas or understand the results? You can monitor usage limits and pull up an interpretation guide programmatically. Connecting through Vinkius means you get full control over your entire content vetting workflow right inside your existing tools. ## Tools ### check_api_health Verifies that the connection to GPTZero is working correctly and ready for analysis. ### detect_ai_in_text Analyzes provided text to determine its likelihood of being generated by a language model. ### get_api_quotas Checks and reports your remaining API credits and usage limits for the billing cycle. ### get_current_user Retrieves basic details about the user authenticated to run the analysis. ### get_interpretation_guide Pulls up a guide that explains what specific scores like perplexity mean in practice. ### get_usage_policy Provides the current API usage rules and rate limit policies for the service. ### list_available_models Lists all supported model versions that can be used for detection jobs. ### submit_prediction_feedback Allows you to send feedback on a previous detection result, helping improve future accuracy. ## Prompt Examples **Prompt:** ``` Analyze this text for AI generation: [insert text]. ``` **Response:** ``` Detection complete! GPTZero identifies a 98% probability that this text was generated by an AI. The perplexity is 12.5 and burstiness is low, indicating highly predictable sentence structures. Would you like a detailed breakdown? ``` **Prompt:** ``` How many API credits do I have left in GPTZero? ``` **Response:** ``` Checking quotas... You have 5,000 character credits and 12 document scans remaining for this billing cycle. Your account is currently on the 'Standard' plan. ``` **Prompt:** ``` Show me the guide on how to interpret perplexity scores. ``` **Response:** ``` Retrieving interpretation guide... Perplexity measures how 'surprised' a language model would be by the text. A score below 10 is usually AI, while human writing typically scores much higher due to creative word choices. Shall I explain burstiness next? ``` ## Capabilities ### Analyze text for AI creation Runs a block of text against proprietary models and returns a confidence score showing the probability that an AI wrote it. ### Measure structural randomness Retrieves perplexity and burstiness scores, which measure how predictable or varied the language structure is. ### Check API usage limits Confirms your remaining character credits and document scan quotas for uninterrupted operation. ### Retrieve technical documentation Accesses the interpretation guide to explain what low perplexity or high burstiness scores actually mean for content integrity. ## Use Cases ### An academic needs to grade a class of papers. Instead of reading every paper manually and guessing if it's human or AI-written, the proctor uses detect_ai_in_text on all submissions. The agent returns a confidence score for each one, allowing them to focus their attention only on suspect texts. ### A marketing team is drafting client copy. The editor runs the first draft through detect_ai_in_text right inside their workspace. The high perplexity and burstiness scores confirm the text has enough unique human variation, ensuring it sounds authentic. ### A compliance team needs to audit a new policy document. The officer uses detect_ai_in_text on sensitive legal texts. If the score is too high, they know the document must be manually reviewed and edited by a human expert before filing. ## Benefits - You instantly verify student submissions using the detect_ai_in_text tool, maintaining academic integrity without leaving your grading dashboard. The confidence score gives you actionable proof of authorship. - The get_api_quotas tool keeps you in control by showing exactly how many credits you have left, so detection operations never fail mid-scan because of an unknown limit hit. - When you don't know what a perplexity score means, the get_interpretation_guide pulls up clear explanations. This knowledge helps you explain complex content metrics to stakeholders accurately. - The system automatically monitors your connection health using check_api_health, letting you know instantly if there's an issue before your agent tries and fails on a massive batch of documents. - You can improve the tool itself by calling submit_prediction_feedback. This sends your real-world validation data back to GPTZero, making future detections more accurate for everyone. ## How It Works The bottom line is that you get automated, quantifiable proof of text origin without ever leaving your main development environment. 1. Subscribe to this MCP and grab your API Key from the GPTZero Dashboard. 2. Connect your AI client through Vinkius, then tell it which text you want analyzed. 3. The agent runs the content through its detectors and returns a detailed score card showing authenticity metrics. ## Frequently Asked Questions **How does GPTZero use detect_ai_in_text? Does it just give a yes/no answer?** No, it provides a confidence score and specific metrics. It doesn't just say 'Yes' or 'No'; it gives you the probability (e.g., 98%) and supporting data like perplexity and burstiness to back up its claim. **What should I check if my detection runs suddenly fail?** You should run check_api_health first. This tool verifies the connection is live. If that passes, then use get_api_quotas to confirm you haven't exceeded your character limits. **Is there a way to track my usage with GPTZero? I need to know my credits.** Yes. Call get_api_quotas anytime. This tool is designed to report your remaining API credits and how many document scans you have left for the billing period. **After using detect_ai_in_text, can I submit feedback to improve accuracy?** Absolutely. Use submit_prediction_feedback to send GPTZero your own judgment on a result. This helps train and make the detection model better for future users. **How do I verify my account details using get_current_user?** It retrieves your profile information instantly. This tool confirms the authenticated user associated with the API key, which is useful for auditing and billing purposes. It tells you exactly whose quota is being consumed. **What models can I use for detection, and how do I see them with list_available_models?** It provides a complete list of all supported model versions available right now. This lets you choose the best balance between speed and accuracy for your specific content analysis needs. You don't have to guess what models work. **How do I understand the rate limits for my API calls using get_usage_policy?** It delivers the official documentation on usage rates and billing rules. This is key for managing large-scale detection jobs, so you know exactly how many requests you can make per minute or hour without getting throttled. **If I get a confusing score, how do I interpret the metrics using get_interpretation_guide?** It pulls up an explanatory guide that breaks down complex concepts like perplexity and burstiness. This helps you translate technical scores into actionable insights about whether content is human-written or AI-generated.