# Yitu Technology MCP > Yitu Technology connects your AI agent directly to enterprise-grade computer vision. This MCP handles everything from real-time facial detection in images to high-precision identity verification and full repository management. It lets you build security workflows that verify identities, read ID cards, and manage biometric data through natural conversation. ## Overview - **Category:** artificial-intelligence - **Price:** Free - **Tags:** computer-vision, facial-recognition, biometric-authentication, identity-verification, image-analysis, security-surveillance ## Description This connector gives your AI agent the ability to run advanced visual intelligence checks on complex datasets. You can move beyond simple image tagging; your agent detects faces in images, verifies if two people pictured belong to the same person using high-precision comparison, and indexes new identities into secure repositories. It’s ideal for any system that relies on verifying who a person is or ensuring data integrity from physical documents. For example, you can detect faces and then use another tool to pull text from an ID card image in one sequence. The real value comes when you build workflows across platforms; because everything runs through Vinkius, your agent can chain this biometric check with other services—like a messaging platform or a database write—all from a single prompt. It ensures that the entire flow is secure and auditable. ## Tools ### detect_active_liveness Verifies if a person is genuinely alive by checking for specific, dynamic actions. ### add_face_to_repo Registers a new face into an existing facial data repository. ### compare_faces Performs a one-on-one comparison to confirm if two faces belong to the same individual. ### create_face_repo Establishes an empty, new facial repository for storing identities. ### delete_face_repo Permanently removes an entire facial data repository and all its contents. ### detect_face Analyzes an image to locate and map the coordinates of faces within it. ### list_repos Retrieves a list of all facial repositories currently managed by the system. ### moderate_image Checks an image for inappropriate or restricted content types before processing it further. ### ocr_id_card Extracts text data from a picture of an official identification card. ### remove_face_from_repo Deletes one or more specific faces from a designated repository. ### search_face_in_repo Searches an entire face repository to find matches for a given face, returning confidence scores. ### detect_silent_liveness Performs anti-spoofing checks that detect if a photo, screen printout, or mask is being used. ## Prompt Examples **Prompt:** ``` Detect faces in this URL: [URL] and tell me how many people are there. ``` **Response:** ``` I've analyzed the image using Yitu Technology. I detected 3 unique faces in the photo. Would you like me to extract the facial locations or perform a comparison against your existing repositories? ``` **Prompt:** ``` Search for this face: [URL] in repository 'REPO_8821'. ``` **Response:** ``` I've performed a search in repository `REPO_8821`. I found a match with 98.2% confidence for Person ID `staff_042`. Should I retrieve the full metadata associated with this person? ``` **Prompt:** ``` List all facial repositories in my Yitu project. ``` **Response:** ``` I've retrieved the list of repositories. You have 2 active collections: 'HQ_Access_List' (ID: `rep_01`) and 'Visitor_Log' (ID: `rep_02`). Would you like me to check the statistics or add a new face to one of these? ``` ## Capabilities ### Identify Faces in Images The agent finds faces in an image, providing their exact location and attributes. ### Verify Identity Match (1:1) It compares two specific images to determine if the individuals pictured are the same person. ### Search Repositories for Matches (1:N) The agent searches a private collection of stored faces and returns the top matching identities with confidence scores. ### Check Against Spoofing It performs advanced liveness detection, verifying that the image or video stream comes from a live person, not a photo or screen. ### Manage Identity Records You can create, list, and delete entire collections of facial data, keeping your identity infrastructure organized. ### Extract Data from IDs The agent reads text off an ID card image, pulling out structured information like names or dates. ## Use Cases ### Onboarding New Staff A new hire needs access control. The agent first uses `ocr_id_card` to pull the name and ID number from their physical card, then runs `detect_face` on a submitted photo, and finally uses `add_face_to_repo` to enroll them into the 'Active Staff' repository. ### Fraud Detection at Checkpoints A checkpoint camera feed is analyzed. The system runs `detect_face`, checks for liveness using `detect_active_liveness`, and then uses `search_face_in_repo` to verify the identity against high-security logs. ### Auditing Data Access Compliance requires knowing who accessed a record. The agent lists all repositories using `list_repos` and verifies if specific individuals are present in each one, reporting back on any discrepancies. ## Benefits - Reduces manual identity checks. Instead of comparing two images manually, you call `compare_faces` and get an immediate boolean match result. - Handles anti-fraud measures automatically. Use `detect_active_liveness` or `detect_silent_liveness` to confirm a person is real before processing any data. - Simplifies record keeping. You can use `list_repos` to see all your active identity collections and then manage them by calling `create_face_repo` when needed. - Streamlines document intake. When processing applications, simply run the image through `ocr_id_card` first; the extracted text is ready for immediate comparison against a face record. - Maintains data hygiene. You can ensure your records are current by running `remove_face_from_repo` when an employee leaves or updates their details. ## How It Works The bottom line is you tell your AI client what security check you need; Vinkius handles running that complex visual logic securely in the background. 1. First, subscribe to this MCP and provide your required Yitu AppID (DevId) and APIKey (DevKey). 2. Second, connect the credentials to your AI client. This ensures all biometric calls run through Vinkius's zero-trust proxy. 3. Third, prompt your agent with a task—for instance, 'Verify this face against my employee database.' The MCP executes the sequence and returns the match result. ## Frequently Asked Questions **How do I use the detect_face tool with Yitu Technology?** You provide an image URL or upload. The agent returns precise coordinates and attributes for every face it finds in that picture. **Can compare_faces verify two different people?** No. `compare_faces` is strictly a one-to-one comparison (1:1). It tells you if the faces match; it cannot tell you how similar they are. **What's the best way to manage my face data with list_repos?** Use `list_repos` first to get all repository IDs. Then, use those IDs when running `search_face_in_repo` to narrow down your search scope. **Does detect_silent_liveness work on old photos?** Yes. `detect_silent_liveness` is designed specifically for anti-spoofing, meaning it can spot deepfakes or printed images used to fool the system. **If a person changes their appearance, how do I use `remove_face_from_repo` to update my records?** You must first call `remove_face_from_repo` using the old face's unique ID. This clears the outdated biometric template from the repository. After removal, you can re-enroll and add their new facial data using a fresh call to `add_face_to_repo`. **Does `moderate_image` check for policy violations or inappropriate content before processing? ** Yes, it acts as an initial safety filter. Before your agent processes the image data, `moderate_image` screens it against predefined policies to flag any prohibited material. This prevents bad data from entering your identity workflow. **What should I do if `search_face_in_repo` returns multiple results with low confidence scores?** When the returned matches have a low confidence score, you need to manually review them. The tool provides the confidence percentage for each match, so you can filter out uncertain identifications and only act on high-certainty data. **What is the proper workflow when I use `create_face_repo` to set up a new identity archive?** `Create_face_repo` establishes the container for your identities, giving you a unique ID. Next, you must populate that repository by calling `add_face_to_repo` repeatedly with all the initial faces. This completes the basic setup process. **How do I find my Yitu AppID and APIKey?** Log in to the [Yitu Cloud Platform](https://www.yitutech.com/), navigate to the 'Developer Center' or 'API Management' section to find your unique AppID (DevId) and APIKey (DevKey). **What is a facial repository?** A facial repository is a secure, private database where you store facial features (templates) of known individuals. This allows the system to perform 1:N searches to identify a person from a large group. **How accurate is the identity comparison?** Yitu is a world leader in facial recognition accuracy. The system returns a confidence score (typically 0.0 to 1.0). A score above 0.8 is generally considered a highly reliable match for the same person.