Build on Sumsub with AI

Sumsub provides a set of tools and resources designed for AI consumption. Whether you are integrating Sumsub into your application with a coding agent, configuring verification flows from natural language, or building autonomous compliance workflows, these capabilities let agents read our APIs, generate code, and act on your account.

Install Sumsub MCP server

The Sumsub MCP server defines a set of tools that AI agents can use to interact with the Sumsub API — creating applicants, generating verification links, inspecting AML cases and answering non-standard questions about your applicants and transactions.

For more connection instructions, see this article.

Install agent skills

Agent skills are packaged instructions that teach AI agents how to accomplish multi-step Sumsub workflows — like setting up a server-side integration, configuring a verification level or querying the Public API.

Install them into your project with agentskills:

npx skills add SumSubstance/agent-skills --all

📘

Note

Requires bash, curl, openssl, and python3 on PATH.

Generate code with OpenAPI spec

A complete OpenAPI 3.x specification is published for the Sumsub Public API. Coding agents and codegen tools consume it directly to produce working SDK calls, request shapes, pagination logic, and error handling without scraping documentation.

Hand the spec URL to your agent at the start of an integration task so generated code matches the current API. The agent will see every endpoint, parameter, and response shape with full descriptions attached.

OpenAPI spec: https://api.sumsub.com/openapi.json.

Plain text documentation

Every Sumsub documentation and reference page is available as Markdown. To fetch the plain-text version, add .md to the end of any docs URL. For example:

• https://docs.sumsub.com/docs/mcp-server.md
• https://docs.sumsub.com/reference/get-fraud-networks.md

Markdown is the format LLMs and agents handle best:

• Lower token cost than HTML.
• Consistent heading hierarchy that agents can parse and navigate.
• Content normally hidden behind tabs or expandable sections is rendered inline.

Paste a markdown URL into any AI chat, or point your agent at it, instead of scraping or copy-pasting from rendered pages.

llms.txt

Sumsub publishes an llms.txt index at the docs root, following the llms.txt convention for pointing AI tools at documentation, API specs, and key resources for a domain.

Most modern AI assistants fetch this file automatically when you reference Sumsub. If you are building your own agent, load https://docs.sumsub.com/llms.txt at the start of a session so the agent knows the layout of our documentation before answering.

Summy AI Copilot inside Dashboard

You can use AI assistance inside the Sumsub Dashboard itself. Sumsub offers Summy AI copilot — an LLM-based assistant to streamline the process of retrieving and visualizing analytical and compliance data.

Where AI fits in Sumsub workflow

AI can support different stages of your Sumsub workflow, from technical integration and Dashboard setup to operational data analysis.

Integration

Engineers can use coding agents (Cursor, Claude Code, Copilot, Codex) to integrate Sumsub into their products. With Sumsub agent skills, the agent reads Sumsub documentation, generates SDK calls, handles webhooks, and troubleshoots errors against live Sandbox data.

Example prompts:

• Add a mandatory KYC step during user registration.
• Write a migration script that moves from our legacy in-house identity check to Sumsub without re-verifying already approved users.
• Write integration tests in Jest that mock Sumsub webhook payloads for approved, rejected, and pending states.

Dashboard configuration

Compliance teams describe the desired setup in plain language and the agent translates it into API calls. You can configure verification levels, AML presets, risk rules, and webhook subscriptions this way without navigating multi-step UI forms.

Example prompts:

• Duplicate our EU KYC level and adapt it for the UK market by adding a proof-of-address step.
• Remove the selfie step from our low-risk onboarding level and replace it with email verification.
• Create a webhook subscription that fires on applicantPending and applicantReviewed events to our new endpoint.

Data exploration

Agents answer non-standard questions about your applicants, transactions, and review queues by treating the Sumsub API as the source of truth.

Example prompts:

• Show me the US users that have been verified today.
• What is the total number of KYC applications this month?
• Why John Smith was rejected?