Yitu Technology MCP. Verify identity and manage facial data via chat.

Yitu Technology MCP Server: Enterprise Biometric Identity Management

Your AI agent gets deep visual intelligence here. This server handles everything from finding faces in an image to verifying complex identities, letting you manage entire biometric workflows with simple commands.

Detecting Faces and Screening Content

You can start by running detect_face, which finds every face in a given image or URL and gives you its exact coordinates. Before any analysis happens, you'll use moderate_image to scan the picture for prohibited material; it flags content that shouldn't be stored or processed.

Verifying Biometric Identity and Searching Databases

The server handles identity confirmation in two ways. For a direct check, run compare_faces. This tool performs a one-to-one comparison to tell you if the two faces provided belong to the same person with high accuracy scores. When you need to search against a large pool of data—say, finding an identity within your private records—you use search_face_in_repo.

This compares a single face against every stored identity in a specific repository, returning all potential matches (1:N) along with confidence levels.

Preventing Spoofing and Ensuring Liveness
The system runs multiple checks to make sure the face is real. You can use detect_active_liveness which requires a physical action from the subject—like blinking or turning—to prove it's a live person, stopping spoofing via movement detection. If you only have a static image, detect_silent_liveness analyzes it for common fraud types, like detecting photos, deepfakes, or 3D masks.

Managing Identity Records and Repositories

Identity data needs structure. You'll start by creating an isolated container using create_face_repo, which sets up a dedicated repository for a specific group of identities. To see what you've got, run list_repos to pull a list of all existing repositories and their current metadata. Once the repo exists, you add data with add_face_to_repo, uploading both the new face image and its associated metadata into that container.

You can also refine your stored identities. If one person leaves or changes records, you use remove_face_from_repo to delete just their specific record while leaving the rest of the repository structure intact. When a group's data is obsolete, you permanently wipe everything using delete_face_repo, which removes the entire repository and all its associated identity data.

Specialized Data Extraction

For identification documents, running ocr_id_card pulls structured text directly from images of ID cards. This extracts names, numbers, dates, and other key fields into usable data points.

This suite lets your AI agent perform complex visual tasks: it detects faces (detect_face), screens content for policy violations (moderate_image), confirms identity via direct comparison (compare_faces), searches across massive private collections (search_face_in_repo), and ensures the person is actually there by checking liveness (detect_active_liveness, detect_silent_liveness).