Model Context Protocol (MCP)

MCP operates on a client-server model with three main elements:

• MCP host. The AI-powered application or agent that initiates requests. This could be a chatbot, an automated compliance workflow, or an identity verification pipeline.
• MCP client. The component embedded within the host that communicates using the MCP standard. It translates the AI model’s needs into protocol-compliant requests.
• MCP server. The external service or data source that receives MCP requests and returns structured results. An identity verification API, a document database, or a sanctions screening service could all function as MCP servers.

MCP acts as the infrastructure layer that makes agentic AI deployable in real-world systems, such as querying databases, running verification checks, and retrieving up-to-date product documentation for faster deployment. That’s because MCP allows agents to safely interact with external systems. 

Autonomous AI agents require secure, structured access to external tools and data. Without MCP, they are either limited or operate in an unstructured and risky way.

Common use cases include:

• For prompts. Structured templates that guide how AI interacts with a system, ensuring consistent and compliant requests.
• For various tools. Executable functions that perform actions, like running checks, submitting data, or updating records.
• For accessing resources. Read-only data sources that AI agents can query, such as files, database records, or verification results.
• For sampling. An advanced feature that enables iterative reasoning, allowing the AI and server to collaborate across multiple steps.

Yes. MCP is an open protocol, meaning its specification is publicly available, and any developer or organization can build MCP-compatible clients, servers, or hosts. The goal is broad ecosystem adoption, so that AI agents across different platforms can connect to the same tools and services without proprietary lock-in.

In terms of Know Your Customer (KYC) or identity verification, MCP enables AI agents to orchestrate multi-step verification workflows, such as document checks, biometric verification, liveness detection, and AML screening. 

This is possible by:

• Connecting to the relevant specialized services through a single, governed protocol.
• Using the AI agent to manage the verification flow without requiring bespoke integrations for each step (and with consistent access controls throughout).