Identity Verification for AI Agents: Didit's MCP Server

Model Context Protocol (MCP) is an open standard for connecting AI agents to external tools and APIs. An identity and fraud check is one of the most consequential tools an agent can call — confirming that a person is real, screening them against sanctions watchlists, or verifying a document before the agent acts on their behalf.

Didit ships a free MCP server today. Any MCP-compatible agent can connect to it and orchestrate the full Didit module catalog — KYC (Know Your Customer), KYB (Know Your Business), AML (Anti-Money-Laundering), biometric verification, document analysis, transaction monitoring — over the same unified /v3/ API used by 1,500+ companies across 220+ countries.

Key takeaways

• MCP is an open standard that lets AI agents call external tools with a structured interface. Didit's MCP server is live and free.
• AI agents need identity verification before they act on behalf of a user: sending money, onboarding a merchant, approving a transaction, or granting access to regulated data all carry fraud and compliance risk.
• Didit's MCP server exposes the full module catalog over the unified /v3/ API. No custom integration per agent, no proprietary wrapper — the same endpoints, the same pricing.
• The MCP server itself has no charge. The underlying module calls are billed at published per-check rates: full KYC core flow $0.33, AML Screening $0.20, Transaction Monitoring $0.02 per transaction.
• 500 free verifications per month apply to checks run via MCP exactly as they do to direct API integrations.
• A dedicated agent-verification product layer is on the near horizon. The MCP server is live and usable now.

What is MCP?

Model Context Protocol is an open standard that gives AI agents a uniform way to call external tools — search engines, databases, APIs, identity services — without each developer writing a custom integration for every service.

An MCP-compatible agent (a Claude-based assistant, a custom LLM agent, a multi-agent pipeline) connects to an MCP server. The server advertises a set of tools with structured schemas. The agent calls those tools with arguments and receives structured results. It reasons over the results, uses them to continue the task, and can call additional tools in sequence.

For an identity service, this means an agent can trigger a KYC session, check its status, retrieve the verification result, run an AML screen on a counterparty, or submit a transaction for monitoring — all programmatically, as part of a larger workflow, without a human writing a new API integration each time.

Why AI agents need identity verification

Agents increasingly perform actions with real financial, legal, or access-control consequences. The same fraud and compliance logic that governs human-facing onboarding applies — often more urgently — when the action initiator is an AI agent.

Onboarding flows. An agent that opens a financial account, registers a merchant, or creates a user profile faces the same KYC obligations as a human-staffed onboarding flow. The agent needs to verify identity and screen against AML (Anti-Money-Laundering) watchlists before completing the action.

Transaction authorization. An agent executing a payment, trade, or fund transfer needs to check the counterparty, apply transaction monitoring rules, and flag anomalies — before the money moves. A real-time decision at $0.02 per transaction is far cheaper than post-hoc remediation.

Access control. An agent granting access to sensitive data, medical records, or regulated content needs a trust signal that the requester is who they claim to be. A biometric check or a liveness verification is that signal.

Ongoing monitoring. Once a user is onboarded by an agent, the same pipeline needs to handle re-verification triggers, AML re-screens when watchlists update, and transaction monitoring alert routing — without a human manually initiating each check.

Without a reliable identity and fraud layer, agents operate on unverified inputs. That is a compliance liability regardless of how sophisticated the model driving the agent is.

How Didit's MCP server works

Didit's MCP server exposes the Didit module catalog as MCP-compatible tools. A connected agent can:

• Create a KYC session and return the session URL for user completion
• Poll or receive webhook events on session status updates
• Retrieve the full verification decision — document data, liveness result, biometric scores, AML status
• Run a standalone AML screen on an individual or entity
• Submit a transaction for real-time monitoring and receive a risk decision
• Run Email Verification or Phone Verification as low-cost pre-KYC filters

All calls route to the unified /v3/ API at https://verification.didit.me. The same endpoints, the same authentication, the same per-check pricing — whether the caller is a backend service or an AI agent orchestrating a complex workflow.

# An agent creates a KYC session via the unified /v3/ API
curl -X POST https://verification.didit.me/v3/session/ \
  -H "x-api-key: $DIDIT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "workflow_id": "your-workflow-id",
    "vendor_data": "agent-run-id-9f2a",
    "callback": "https://your-agent-backend.com/webhook"
  }'

The response carries a session_url for the user to complete (in a browser or embedded flow) and a session_id the agent tracks. When verification completes, the webhook fires and the agent continues its task with the verified result in hand.

Agentic verification patterns

Step-up verification in a chat agent. A financial assistant handles account queries without friction for low-risk requests. When the user asks to transfer funds above a threshold, the agent triggers a biometric re-authentication via Didit's MCP server and holds the action until the check clears. No hard-coded rule in the UI — the agent makes the decision and the identity check is the enforcement mechanism.

Merchant onboarding agent. A multi-agent pipeline collects business registration details, triggers KYB with UBO extraction, and automatically spawns linked KYC sessions for each beneficial owner — producing a closed-loop compliance record without a human coordinating each step. The same Didit workflow that a human-facing console would run is orchestrated entirely by the agent over MCP.

Compliance co-pilot. An internal agent monitors the transaction feed. When a Transaction Monitoring alert fires via webhook, the agent runs AML screening on the flagged party, drafts an investigation summary, and routes the case to an analyst — without a human initiating each step of the triage.

Use cases

Fintech product agents. Agents that handle account opening, fund transfers, or portfolio adjustments need KYC at onboarding and transaction monitoring at every payment. MCP makes both available without hard-wiring a new verification integration into each agent build.

Crypto and Web3 agents. Agents executing trades, bridging assets, or managing wallet operations need AML screening and wallet risk scoring on counterparties. BYOK (bring-your-own-key) wallet screening at $0.02 per check scales with agent-driven transaction volume without expensive annual contracts.

Marketplace automation. Seller-onboarding agents can trigger KYB plus linked KYC for new merchants, run database validation, and return a verified status — before a human account manager reviews the account. The agent front-loads the compliance work.

Agentic commerce. Agents acting on behalf of a user in a shopping or booking context need to confirm the user is real and authorized before completing a purchase or reservation. Identity verification becomes a trust signal the agent carries into every transaction it executes.

How Didit helps

Didit's MCP server is free and live now. Connect any MCP-compatible agent — point it at the Didit MCP server, supply your API key, and the full module catalog is available as agent-callable tools. Documentation is at docs.didit.me.

The underlying modules are billed at published per-check rates: KYC core flow $0.33, AML Screening $0.20 per screen, Transaction Monitoring $0.02 per transaction, Wallet Screening from $0.02 BYOK. 500 free verifications per month. No minimums.

Coverage: 220+ countries, 14,000+ document types, 48+ languages, sub-2-second inference. Certifications: SOC 2 Type 1, ISO/IEC 27001:2022, iBeta Level 1 PAD. Only provider formally attested by an EU member-state government (Spain's Tesoro / BdE / SEPBLAC / CNMV) as safer than in-person verification.

A dedicated agent-verification product layer — with trust signals and controls designed specifically for agentic workflows — is on the near horizon.

Frequently asked questions

Is Didit's MCP server free?

Yes. The MCP server has no charge. You pay for the underlying module calls at published per-check rates: KYC core flow $0.33, AML Screening $0.20, Transaction Monitoring $0.02 per transaction. 500 free verifications per month apply.

What agents are compatible with Didit's MCP server?

Any agent that supports the Model Context Protocol standard — Claude-based agents, custom LLM agents built on agent frameworks, and multi-agent pipelines. The server is framework-agnostic.

Does using MCP change how verification results are returned?

No. The MCP server calls the same /v3/ API endpoints as a direct integration. Results are identical — full session decisions, AML responses, transaction statuses, and webhook events.

Is there a dedicated agent-verification product?

A dedicated agent-verification product layer is on the near horizon. The MCP server is live and usable today for agent-orchestrated verification over the unified /v3/ API.

Can I test this before committing?

Yes. Sign up at business.didit.me for 500 free verifications per month, connect the MCP server, and run the full module catalog without a contract.

Ready to get started?

• Read the docs → docs.didit.me — MCP server setup, API reference, module catalog
• Explore the platform → didit.me/products/id-verification
• See pricing → didit.me/pricing — every module, public price, no minimums
• Start free → business.didit.me — 500 free verifications / month, MCP server included