Use the MCP server to let AI agents perform Sumsub operations on your behalf, while keeping access controlled by your existing Sumsub role permissions.

How Sumsub MCP server works

Sumsub supports the Model Context Protocol (MCP), which allows AI agents to interact with your Sumsub account through a single server URL. After you add the MCP server to your client and authenticate, your agent can perform the actions allowed by your Sumsub role permissions.

With the Sumsub MCP server, you can create agent-based workflows for onboarding, compliance operations, transaction monitoring, and configuration discovery. 

You can connect MCP-compatible AI agents, such as Claude Desktop, Claude Code, ChatGPT, Codex, Cursor, or your own agent, directly to your Sumsub account.

Connect MCP server

Before you start, make sure that:

• You have an active Sumsub account.
• Your role has the Use MCP server permission enabled.

Use https://api.sumsub.com/mcp/ as base URL to connect to the Sumsub MCP server.

To work with test data, use the Sandbox base URL https://api.sumsub.com/mcp/sandbox/ instead. For details, see Test in Sandbox.

Connect ChatGPT

To connect ChatGPT:

1. Open official Sumsub app for ChatGPT.
2. Click the Connect button.
3. Select the Sign in with Sumsub option.

Connect Claude

To connect Claude:

1. Open official Sumsub Connector for Claude.
2. Click the Connect button.

Connect Claude Code

To connect Claude Code:

1. Open Terminal.

2. Add the Sumsub MCP server: claude mcp add --transport http sumsub https://api.sumsub.com/mcp/

3. Run Claude Code.

4. Type /mcp. 

5. Select the sumsub MCP server and authenticate.

Connect OpenAI Codex

To connect OpenAI Codex:

1. Open Terminal.
2. Add the Sumsub MCP server: codex mcp add sumsub --url https://api.sumsub.com/mcp/.
3. Paste the following to authenticate: codex mcp login sumsub.

Connect Cursor

To connect Cursor:

1. Open the .cursor/mcp.json file.
2. Add the following configuration:
   JSON

   {
    "mcpServers": {
      "sumsub": {
        "command": "npx",
        "args": [
          "mcp-remote",
          "https://api.sumsub.com/mcp/"
        ]
      }
    }
   }

Test in Sandbox

Use the Sandbox MCP server to try agent workflows on test data before you connect your agent to Production. It provides the same tools as the Production MCP server but works only with your Sandbox data: the system does not perform real checks, and no billing is generated.

Use https://api.sumsub.com/mcp/sandbox/ as base URL to connect to the Sumsub MCP server in Sandbox.

You can keep both servers connected at the same time under different names, for example sumsub and sumsub-sandbox, and tell your agent which one to use.

Connect Claude to Sandbox

1. In Claude, go to Settings → Connectors.
2. Click Add custom connector.
3. Enter a name, for example Sumsub Sandbox, and the URL https://api.sumsub.com/mcp/sandbox/.
4. Click Add, then click Connect and sign in with your Sumsub account.

Connect Claude Code to Sandbox

1. Open Terminal.
2. Add the Sumsub Sandbox MCP server: claude mcp add --transport http sumsub-sandbox https://api.sumsub.com/mcp/sandbox/
3. Run Claude Code.
4. Type /mcp.
5. Select the sumsub-sandbox MCP server and authenticate.

Connect OpenAI Codex to Sandbox

1. Open Terminal.
2. Add the Sumsub Sandbox MCP server: codex mcp add sumsub-sandbox --url https://api.sumsub.com/mcp/sandbox/.
3. Paste the following to authenticate: codex mcp login sumsub-sandbox.

Connect Cursor to Sandbox

1. Open the .cursor/mcp.json file.
2. Add the following configuration:
   JSON

   {
    "mcpServers": {
      "sumsub-sandbox": {
        "command": "npx",
        "args": [
          "mcp-remote",
          "https://api.sumsub.com/mcp/sandbox/"
        ]
      }
    }
   }

Send feedback

If something you need is missing from the Sumsub MCP server — a tool, parameter, or workflow — just ask your agent to send us feedback.

The agent will use the built-in feedback tool to share your suggestion with the Sumsub team, and we use this input to prioritize new capabilities.

Example request: Send feedback to Sumsub: I would like the MCP server to support bulk applicant import.

Request examples

The following examples show how you can ask your agent to perform common Sumsub operations and which MCP tools the agent can call to complete each request.

Self-service onboarding from chat

Ask your agent to onboard an applicant and send a verification link.

Example request: Onboard John Doe at the Standard KYC level and send him the verification link.

The agent can call:

• applicant_create_individual to create the applicant.
• verification_link_create to generate the verification link.

Compliance-team copilot

Ask your agent to check an applicant current status and review history.
Example request: What is the current status of applicant A12345, and what triggered the last review?

The agent can call:

• applicant_get to retrieve applicant data.
• applicant_timeline_get to inspect applicant events.
• applicant_moderation_state_list to check moderation states.

Transaction-monitoring triage

Ask your agent to retrieve transactions that triggered monitoring rules.

Example request: Show me transactions for applicant A12345 over the last 7 days that hit a monitoring rule.

The agent can call:

• transaction_list with filters.
• transaction_get to retrieve transaction details.

Configuration discovery

Ask your agent to inspect configured verification levels.

Example request: What verification levels do we have configured, and how do they differ?

The agent can call:

• verification_level_list to retrieve available levels.
• verification_level_get to inspect selected levels.