Build with AI

Connect Prynt to your AI assistant

The Prynt MCP server gives Claude, Cursor and other MCP clients direct access to the Prynt API — with your key, on your machine.

The Model Context Protocol (MCP) lets an AI assistant call tools you run locally. The Prynt MCP server (@pryntid/mcp) exposes the Prynt server API as tools, so your assistant can read real identification events while it helps you write the integration — instead of guessing the response shape.

01 Nothing to install by hand

You need Node.js 18 or newer. The server is not on the npm registry yet, so pryntid.com serves the package and npx fetches it on first run — this is the command your assistant will start:

npx -y --package=https://pryntid.com/downloads/prynt-mcp-0.1.0.tgz prynt-mcp

02 Connect your assistant

Use a secret key from the console. Start with a test key (sk_test_…): it only reads and writes the test environment.

Claude Code:

claude mcp add prynt -e PRYNT_SECRET_KEY=sk_test_your_secret_key \
  -- npx -y --package=https://pryntid.com/downloads/prynt-mcp-0.1.0.tgz prynt-mcp

Claude Desktop (claude_desktop_config.json) and Cursor (~/.cursor/mcp.json or .cursor/mcp.json) use the same block:

{
  "mcpServers": {
    "prynt": {
      "command": "npx",
      "args": ["-y","--package=https://pryntid.com/downloads/prynt-mcp-0.1.0.tgz","prynt-mcp"],
      "env": {
        "PRYNT_SECRET_KEY": "sk_test_your_secret_key"
      }
    }
  }
}

The server talks to https://api.pryntid.com by default; "PRYNT_ENDPOINT" in env overrides that base URL.

03 Ask it

Read the Prynt integration guide, then add identify-on-signup to this form
and verify the requestId in our signup route with the secret key.

Look up Prynt request req_… — does this signup look risky? Does the device
already have other accounts?

Check 203.0.113.7 with Prynt: is it a datacenter, VPN, Tor or proxy address?

Is the Prynt API healthy?

Tools

  • prynt_get_event — GET /v1/events/{requestId}: the server-side record of an identification (visitorId, confidence, decision, riskScore, Smart Signals, IP and location, accountsOnDevice). Read-only.
  • prynt_lookup_ip — POST /v1/ip: country, ASN and datacenter / VPN / Tor / proxy flags for a raw IP. Read-only; does not use identification quota.
  • prynt_health — GET /v1/health: API status per component. Read-only.
  • prynt_integration_guide — install @pryntid/js, identify in the browser, verify on your server. No API call.
  • prynt_report_outcome — POST /v1/outcomes: writes a label such as fraud or legit. Bad labels flag that device, account or IP on future visits in your account and, on Prynt's cloud, also feed the shared reputation network as one-way hashes unless you turned contribution off in the console (Rules › Reputation network). Approve each call yourself.

Security

  • The server runs on your machine. Your secret key is read from its environment and sent only to the Prynt API; no tool returns it.
  • Tool results become part of your conversation with the assistant. Events include personal data such as IP addresses and user agents, so prefer test data.
  • Never commit a config file that contains a key.

Prefer to integrate by hand? The docs cover install, identify and verify in three steps.