Facesign MCP

What facesign-mcp MCP server does

The facesign-mcp MCP server is intended for building and testing FaceSign step-up verification flows from an AI coding tool. FaceSign sessions are represented as directed graphs called flows. A flow must contain one START node and at least one END node, while intermediate nodes perform user-facing or validation tasks.

Flows can support several session purposes, including identity verification, collecting data, authorization, and analysis. Nodes connect through outcomes. Each outcome points to another node, so every declared route must lead somewhere rather than remain unlinked.

How it works

The Node.js client is initialized with a FaceSign integration token, typically read from the FACESIGN_TOKEN environment variable. After initialization, the client can make requests to FaceSign API endpoints.

A basic flow begins at a START node and finishes at one or more END nodes. Conversation nodes can either provide exact speech for the avatar or describe a goal for a multi-turn exchange. Their conditional outcomes use natural-language criteria to select the next node. Goal-driven conversations should define practical exit conditions, such as repeated non-response, a maximum number of exchanges, or refusal to continue.

Permission nodes can request camera and microphone access and route separately for granted and denied results. Liveness detection requires camera access and several seconds of video. The documented recommended order is PERMISSIONS, CONVERSATION, then LIVENESS_DETECTION; placing liveness detection immediately after a permission request does not provide enough time to accumulate video.

Setup and configuration

Install the FaceSign Node.js package with either npm or Yarn:

• npm install @facesignai/api
• yarn add @facesignai/api

Create a client with the integration token:

const { Client } = require("@facesignai/api")
const facesignClient = new Client({
  auth: process.env.FACESIGN_TOKEN,
})

The supplied material identifies FACESIGN_TOKEN as the authentication input. It does not specify an MCP client configuration, transport, or additional environment variables.

Tools and capabilities

The facesign-mcp MCP server material documents capabilities for constructing flows with these node types:

• START and END nodes for flow boundaries
• CONVERSATION nodes for direct speech or goal-oriented dialogs
• PERMISSIONS nodes for camera and microphone access
• LIVENESS_DETECTION nodes for real-person, deepfake, and no-face outcomes
• ENTER_EMAIL nodes for collecting an email without verification
• DATA_VALIDATION nodes for checking collected fields
• DOCUMENT_SCAN nodes using Microblink-powered identity-document scanning

Document scanning can extract information such as a name, date of birth, and document number into the session report. For email collection with OTP verification, the documentation directs users to TWO_FACTOR_EMAIL rather than pairing ENTER_EMAIL with another step.

Limitations and notes

The provided material describes the FaceSign SDK and flow model but does not enumerate named MCP tools, supported MCP transports, client compatibility, or a license. It also indicates that permission requests happen automatically when a flow has no PERMISSIONS node, while an explicit permission node is useful for custom prompts or separate denial handling. A conversation node marked doesNotRequireReply: true can deliver a final message before transitioning to an END node.