MCP for KYB: Verify Businesses and Beneficial Owners with AI Agents

Know Your Business (KYB) is the hardest identity workflow to automate because it fans out. You start with one company, pull its registry record, list its officers and Ultimate Beneficial Owners (UBOs), then have to verify each of those humans and screen everyone — the entity and every person — against sanctions, politically exposed person (PEP), and adverse-media data. Stitching that together by hand is where most integrations stall. Didit's official Model Context Protocol (MCP) server turns the whole chain into a conversation. An AI agent connected to https://mcp.didit.me/mcp can run registry lookup, resolve officers and UBOs, spin up a linked KYC session for each owner, and screen the entity plus each person for Anti-Money Laundering (AML) risk — a closed-loop KYB flow that no other provider offers as one composable path.

Key takeaways

• Didit's MCP server exposes 130+ tools across 11 categories over Streamable HTTP at https://mcp.didit.me/mcp, including full KYB session tooling.
• An agent runs the KYB chain in order: registry lookup → officers + UBOs → a linked KYC session per UBO → entity + person AML → risk scoring — the closed loop is unique to Didit.
• Authentication is OAuth 2.1 + Proof Key for Code Exchange (PKCE) via business.didit.me — no API key on the hosted server, and access respects your console role.
• Each UBO gets a real KYC session (ID document, liveness, face match), so beneficial ownership is verified against live people, not just registry names.
• KYB starts at $2.00, AML screening is $0.20, and the linked KYC sessions bill at the standard $0.33 — with 500 free verifications a month and the MCP layer free.
• Connect Claude Code in one line: claude mcp add --transport http didit https://mcp.didit.me/mcp.

What KYB over MCP actually does

The Model Context Protocol is an open standard for letting AI agents call external tools consistently. Didit's server is the official MCP implementation for its identity and fraud platform, published under the MIT license at github.com/didit-protocol/mcp and served over Streamable HTTP at https://mcp.didit.me/mcp. Among its 130+ tools across 11 categories, the Sessions (KYC/KYB), Standalone Verification, Transaction Monitoring & AML, and Vendor Businesses categories are the ones a business-verification flow leans on.

The point of running KYB through MCP is that the agent orchestrates the fan-out for you. Instead of calling a registry endpoint, parsing the officer list, deciding which shareholders cross the beneficial-ownership threshold, creating a KYC session for each, and firing AML checks, you describe the outcome and let the agent walk the chain.

The closed loop: registry to UBO to AML

Here's the sequence an agent runs when you ask it to verify a business:

1. Registry lookup. The agent queries the official company registry for the entity — legal name, registration number, status, and address.
2. Officers and UBOs. It resolves the directors and officers, then the Ultimate Beneficial Owners — the real humans who ultimately own or control the company.
3. Linked KYC per UBO. For each UBO, the agent creates a linked KYC session. That owner completes an ID document check, passive liveness, and face match — so ownership is verified against a live person, not just a name on a filing.
4. Entity + person AML. The agent screens the company itself and each officer and UBO against sanctions, PEP, and adverse-media data.
5. Risk scoring. The results roll up into a single risk view for the business.

This is the closed loop that's unique to Didit's business verification: KYB spawns real KYC sessions, which feed AML, which feeds a risk score — one composable pipeline rather than five stitched vendors.

Running it in natural language

Once the MCP server is connected, the whole chain is a prompt. Something like:

"Run a full KYB check on this company: pull the registry record, identify the officers and UBOs, send each UBO a KYC verification link, screen the entity and every person for AML risk, and give me a summary with the risk score when everything's back."

The agent creates the KYB session, walks the registry-to-UBO chain, hands you the KYC links to distribute (or delivers them however your onboarding email flow expects), polls each session, runs the AML screening, and reports a consolidated result. You can follow up conversationally: "Which UBOs still haven't completed their KYC?" or "Show me only the officers with an AML hit." Because the tools are composable and user-scoped, the agent stays inside your console permissions the whole way.

Authentication and access control

The hosted server uses OAuth 2.1 with PKCE — the "Log in with Didit" flow through business.didit.me. There's no API key to manage for the hosted setup. Access is user-scoped across two scopes, didit:management and didit:verification, and every action the agent takes respects your existing console role. If your role can't create businesses in the UI, the agent can't either. For KYB — where you're touching sensitive ownership and screening data — that inherited permission boundary matters.

Connecting your client

Claude Code connects in one command:

claude mcp add --transport http didit https://mcp.didit.me/mcp

Run /mcp to complete the "Log in with Didit" OAuth flow, and the KYB tools appear in your session. Claude Desktop, Cursor, VS Code, Windsurf, and Zed use a JSON config pointing at the same Streamable HTTP URL:

{
  "mcpServers": {
    "didit": {
      "type": "http",
      "url": "https://mcp.didit.me/mcp"
    }
  }
}

ChatGPT Developer Mode is supported through OpenAI's beta connector; since that path is still in beta, its exact steps may change. Full client instructions are in the MCP overview docs.

Pricing and screening scope

Business verification starts at $2.00, AML screening is $0.20 per check, and each linked UBO KYC session bills at the standard $0.33 for the full document, liveness, and face-match bundle. Every account gets 500 free verifications a month, and the MCP layer adds nothing — you pay per successful check regardless of where it's triggered.

On screening: Didit checks entities and people against sanctions lists, PEP databases, and adverse-media sources — described generically because the point is coverage, not any single data feed. The AML result is part of the same closed loop, so the agent can factor it straight into the risk score it reports back.

Didit is Y Combinator W26, profitable, has raised $7.5M, and serves 1,500+ companies in production across 220+ countries and 14,000+ document types — so a KYB flow you prototype over MCP scales to real onboarding volume without re-plumbing.

Start free

Connect an AI agent to https://mcp.didit.me/mcp and run a full business verification — registry, UBOs, linked KYC, and AML — from a single prompt. Create your workspace and start free with 500 verifications every month, no card required. Read the MCP overview, browse the open-source server, or explore the developer hub to see the full KYB toolset.