Goodlabs
Connect over MCP

Bring Goodlabs to your AI

Connect your Goodlabs health data to Claude, ChatGPT, Codex, Grok, or any MCP client over the Model Context Protocol (MCP).

Once connected, your assistant can:

  • Read biomarkers & trends

    Exact values, units, and history for any marker you track.

  • Track health scores

    How each score is built and how it has moved over time.

  • Check orders & results

    Order status and what is still in flight, in plain language.

  • Book or cancel visits

    Find nearby lab appointment slots and schedule from chat.

  • Explore tests & build a cart

    Search the Goodlabs catalog and queue up new tests to add.

  • Back it with research

    Pull relevant PubMed studies into the conversation.

Already use our in-platform AI, Barry? You can continue to chat with Barry in the Goodlabs platform. Barry runs state-of-the-art models and is tuned for working with your health data. The option to connect over MCP is for those who’d rather use their own AI assistant.

Goodlabs MCP URL

https://goodlabs.com/mcp/user

Claude Web walkthrough

Claude on the web at claude.ai.

  1. Open the custom connectors page

    Open Claude connectors

    Go to the link directly, or in Claude open Settings → Connectors and choose "Add custom connector".

  2. Paste the Goodlabs MCP URL

    Drop the endpoint below into the URL field and continue.

    MCP server URL

    https://goodlabs.com/mcp/user

  3. Sign in to Goodlabs

    Claude opens a Goodlabs sign-in window. Approve access and your data is connected, and your tools show up in the chat composer.

Why connect Goodlabs?

Structured, AI-ready data

Goodlabs cleans up, standardizes, and labels every result. An LLM reads it in context (units, reference ranges, history) instead of guessing from a messy PDF.

One home for your health

The more of your bloodwork and history lives in Goodlabs, the more your assistant can actually work with. Upload past results and add new tests to keep it whole.

Use the model you like

Prefer Claude, Grok, or Codex over another chatbot? Connect Goodlabs and ask there. Barry stays available in-app whenever you want a guided read.

Building your own client or agent?

Point any MCP client at https://goodlabs.com/mcp/user. Goodlabs supports OAuth 2.1 with dynamic client registration, PKCE (S256) and no client secret, so there is nothing to apply for.

  • Running locally? Use a loopback callback such as http://localhost:<port>/callback (127.0.0.1 and [::1] work too, any port and path). It registers instantly and shows up on the consent screen as your own computer.
  • Hosted somewhere else? Any https://callback is accepted at registration. Because Goodlabs can't verify who runs it, the consent screen labels the app “Unrecognized”, shows the callback address, and asks the person connecting to confirm they trust it before anything is shared.
  • Not accepted: plain http:// to a LAN IP or tunnel hostname, callbacks with a URL fragment, and custom URI schemes. Switch to loopback or HTTPS, or email support@goodlabs.com with your client name and exact redirect URI if your app genuinely can't.

Every connection is scoped to the signed-in member's own data, and checkout for any tests your agent adds to the cart always happens on goodlabs.com.

More data in, smarter answers out

Upload past lab results and book new tests so your assistant has the full picture. Your data stays in Goodlabs; you just choose where to read it.

Get started