ProofAge MCP server: let your AI coding agent integrate KYC and age verification

The ProofAge MCP server lets an AI coding agent such as Claude Code or Cursor integrate identity (KYC) and age verification into your app. Connect it at https://app.proofage.net/mcp/admin and sign in with OAuth, with no API keys. The agent creates a test workspace, writes the code with our SDKs and tests the webhooks. It is free in every account.

Start free: 500 verifications

Connect in one step

Your MCP client opens a browser window. You sign in with your ProofAge console account and approve access. There are no API keys to paste.

Claude Code

claude mcp add --transport http proofage https://app.proofage.net/mcp/admin

Then run /mcp in Claude Code and choose proofage to sign in.

Claude.ai

In Settings → Connectors, add a custom connector with the URL https://app.proofage.net/mcp/admin, then connect it and sign in.

Cursor

In ~/.cursor/mcp.json:

{
  "mcpServers": {
    "proofage": { "url": "https://app.proofage.net/mcp/admin" }
  }
}

VS Code

In .vscode/mcp.json:

{
  "servers": {
    "proofage": { "type": "http", "url": "https://app.proofage.net/mcp/admin" }
  }
}

Grok

On grok.com, open Connectors → New connector → Custom, enter https://app.proofage.net/mcp/admin, then sign in. On Business and Enterprise plans, an admin may have to add the connector first.

Perplexity

In Settings → Connectors, add a custom remote connector with the URL https://app.proofage.net/mcp/admin and OAuth authentication, then sign in. Custom connectors need a Pro, Max or Enterprise plan.

ChatGPT, Codex and other clients

Point the client at https://app.proofage.net/mcp/admin. The server supports OAuth discovery and dynamic client registration. ChatGPT, GitHub Copilot in VS Code, Gemini CLI, the Gemini app, Mistral Le Chat, Copilot Studio, Superhuman and Coda, Amazon Quick, n8n Cloud, and clients that sign in through a local callback, such as Codex, can connect. If your client is refused at sign-in, write to us with its name: we add clients on request.

MCP server in the docs

What your agent can do

The agent works with your account the way your team does in the console. Plain requests work: the agent picks the tools.

The four groups of MCP tools, with an example prompt for each
Group What it does Example prompt
Integrate Create a test workspace, read its keys and URLs, create verifications, set test outcomes and replay webhooks against a local handler. Then create the live workspace. “Integrate ProofAge age verification into this app. Use a test workspace, the Browser SDK on the checkout page, and a webhook handler that marks the user as verified.”
Investigate Look up verifications, their decision and reason, images, webhook deliveries and statistics, with the same data as the console. “Why was verification 550e8400-… declined?”
Support Send a verification link, block or unblock a face, and erase a person's data on request. “Send a verification link for customer 8812 in the "Checkout" workspace and give me the URL.”
Docs The agent reads the developer docs and the OpenAPI spec through the server, with a tool or as MCP resources, so it writes code against the real API. “Check our webhook handler against the ProofAge webhooks guide: verify the signature with the SDK, and skip a delivery whose X-ProofAge-Webhook-Delivery-Id we have already processed.”

More example prompts · MCP tools reference

From zero to live, step by step

With the SDKs, a typical integration is built and tested within a day. With an AI coding agent using the MCP server, it typically takes a couple of hours. This is the order the agent follows.

  1. 1

    Create a test workspace

    With the check you need: Identity verification (KYC), Facial age check or Age verification by ID. Test mode is free.

  2. 2

    Read the integration guide

    The agent gets the workspace's keys and URLs, and reads the quick start, the SDK guides, webhooks and testing from the docs.

  3. 3

    Write the integration with the official SDK

    For Node.js, Python, PHP or Laravel. The SDK signs requests and verifies webhooks, so the agent does not write its own signing code.

  4. 4

    Create a verification and finish it yourself

    The agent gives you the link. The camera steps need a person, so you complete them.

  5. 5

    Set the test outcome

    Approved, declined or resubmission requested. The outcome sends the webhook to your handler.

  6. 6

    Replay webhook deliveries

    The agent fetches the exact signed request and sends it to your local handler until it passes.

  7. 7

    Go live

    The agent can create the live workspace. You copy its secret key from the console and put it in production: the live secret key is never shown to the agent.

The integration prompt

A ready brief to paste into your agent. It starts like this:

Integrate ProofAge into this project.

What I want: <e.g. an age check (18+) before checkout; or identity verification (KYC) for new sellers>.
Where the person starts it: <e.g. the "Verify age" button on /checkout>.
What a result should change: <e.g. set users.age_verified_at on approval; block the order on decline>.

The full integration prompt

Built to stay safe

The agent acts as you, in the account you work in, and sees only the tools your console role allows. Every action is recorded under your name, as it is in the console.

  • Live secret keys stay with people

    They are never returned through MCP. A person copies them from the console.

  • Live changes wait for you

    A change to a live workspace or its keys is only a preview. Nothing is applied until you confirm.

  • Destructive actions take two steps

    Blocking or unblocking a face and erasing a person's data first return a preview and a short-lived token. The agent then confirms with an exact phrase and a reason.

  • Roles and rate limits

    Developers cannot erase personal data. Creating verifications, workspaces and test outcomes, and erasing data, are rate limited per user.

After launch

Support and operations teams keep using the same server from Claude.ai or their own assistant, with the support specialist role.

  • “Why was this user declined?” The agent reads the decision, its reason and what the person submitted.
  • “Send this customer a verification link.” The agent creates a verification and gives you the link to send.
  • “Erase this person's data.” The agent prepares the erasure, shows what will be removed and what stays, and confirms once you agree.

MCP tools reference

Questions about the MCP server

Does ProofAge have an MCP server?

Yes. The ProofAge MCP server is a remote MCP server over HTTP at https://app.proofage.net/mcp/admin. You sign in with your ProofAge console account over OAuth. An AI coding agent can use it to integrate identity (KYC) and age verification into your app, look up verifications and handle support requests. See also the KYC API and the developer overview.

Which AI coding agents work with it?

The docs have setup steps for Claude Code, Claude.ai, Cursor, VS Code, Grok and Perplexity. ChatGPT, Codex, GitHub Copilot in VS Code, Gemini CLI, Mistral Le Chat and several other clients connect with the same URL. If your client is refused at sign-in, write to us: we add clients on request.

Do I need an API key to connect?

No. Your MCP client opens a browser window, you sign in with your ProofAge console account and approve access. The server supports OAuth discovery and dynamic client registration.

Can the agent touch live data or keys?

Only within your console role, and with guards. Live secret keys are never returned through MCP: a person copies them from the console. A change to a live workspace or its keys is only a preview until you confirm it. Blocking or unblocking a face and erasing a person's data need a second confirmation step with a short-lived token, an exact phrase and a reason. Every action is recorded under your name.

How long does an integration take with an AI agent?

With the SDKs, a typical integration is built and tested within a day. With an AI coding agent using the MCP server and the integration prompt, it typically takes a couple of hours.

Is the MCP server free?

Yes. The MCP server is included in every account at no extra cost. Live verifications it creates are billed as usual, and test mode is always free. See pricing.

A direct line to the developers who built it

Write to us on LinkedIn, by email or in the dashboard chat. The developers who built ProofAge answer themselves, weekends included.

Connect your agent and start building

Test mode is free. 500 live verifications are free too, no card, no time limit.