Ours Privacy MCP Server

Picture asking your assistant, "Why did booked appointments drop last week?" and getting an answer drawn from your own Insights, Funnels, and Journeys. Or asking it to "set up a cookie banner for example.com and publish it" and watching it do the work. The Ours Privacy MCP server makes that possible by connecting an AI assistant such as Claude, ChatGPT, or Codex to your account through the Model Context Protocol.

Connect it once, then talk to your account the way you would talk to a teammate. Anything the assistant does counts as your action, so connect only assistants you trust.

https://ai.oursprivacy.com/mcp

Tip: The built-in AI Assistant gives quick answers about the screen you are on, right inside Ours Privacy. The MCP server brings your whole account into the assistant or IDE you already use, to ask questions and to build.

What you can ask and build

Ask about anything in your account, then ask the assistant to do the work. It reaches every part of Ours Privacy, from the first web source to the published version:

• Understand what happened. "Which campaign drove last week's appointment requests, and where did people stop in the booking funnel?" or "Build a monthly dashboard with a highlights note, my booking funnel, and the ten most recent booking sessions." It reads your Insights, Funnels, Journeys, attribution, and dashboards, explains what it finds, and assembles dashboards with notes, session replay lists, and your report logo.
• Find friction. "Which pages had the most rage clicks this month? Give me a few replays to watch." It ranks heatmap pages by friction and hands back recorded sessions for you to review.
• Read consent and campaign results. "Show opt-out rates by region for my cookie banner, then break California down by page" or "How many scans did the spring flyer QR code get, and from which cities?"
• Collect the right data. "Create a web source for example.com, give me the install snippet, and confirm events are arriving." It also manages allowed events and data governance, so you decide which events and categories move on.
• Send it where it needs to go. "Create a Klaviyo destination called Main Klaviyo, map my purchase event to it, and check that it is delivering." Field mappings, conditions, and the fallback mapping are all in reach.
• Manage tags. "Add a Tag Manager tag that fires on the thank-you page, put it in the Conversions folder, and give me the install snippet." Tags, triggers, variables, and folders work the way they do in the app.
• Govern consent and privacy. "Create a cookie banner for example.com with an opt-in rule for California," "Record sessions only on app.example.com," or "Which trackers from the last web scan are not covered by consent?" It updates consent categories and regional rules, session replay recording, and scanner suppressions after you review them.
• Experiment and personalize. "Create Pricing v2 for half of mobile visitors and start it," then later "How is it doing? Stop it and declare variant B the winner." It handles targeting, variants, goals, live results, and the visitor traits that drive personalization. See Manage Experiments with AI.
• Reach more visitors. "Create a short link with campaign UTMs for the spring flyer" or "Add a language selector to example.com."
• Ship safely. "What is in the draft?" "Publish it." "Roll back to yesterday's version." Every change lands in a draft, so you can see the difference before it goes live and return to any earlier version.

You never name a tool. The assistant works out what it needs from the catalog of Ours Privacy capabilities, and common multi-step jobs (a consent banner, a destination end to end, a new web source, a web scanner audit) follow guided walkthroughs.

The assistant answers within your permissions, and nothing reaches your site until you publish or start an experiment.

Check changes to your site

You can ask your assistant to rescan your site or capture fresh evidence from a single page after changing a tracker or consent setup. Follow verification progress, review capture results, and revisit earlier verifications in the same conversation.

Page verification lets you check a change while keeping your site's full-scan history and schedule intact.

Connect with OAuth

OAuth is the simplest way in: no API key to create, copy, or rotate. The assistant acts as you, with your current permissions in the organization you choose.

Before you start: an organization admin turns on OAuth connections under Settings → Account Settings. It is off by default.

In Claude Desktop or claude.ai:

1. Open Customize → Connectors and select Add connector → Add custom connector.
2. Enter a name, set the URL to https://ai.oursprivacy.com/mcp, and select Continue.
3. Keep Sign in now and Use Claude's published identity selected, then select Add.
4. On the connector page, select Connect, sign in to Ours Privacy, pick the organization, and select Allow.

Claude asks for approval the first time it uses each capability in a chat. Access ends when you disconnect in Claude, or when an administrator turns off OAuth connections, removes the member, or changes their permissions. Those changes take effect immediately. To give an assistant less access than you have, connect it with a read-only API key instead.

Full reference: Claude Connectors docs.

Connect with an API key

ChatGPT, Codex, Claude Code, and Cursor connect with an API key. Create one on the API Keys page (see Platform API keys), and choose the read-only preset when the assistant only needs to look.

Claude Code

claude mcp add --transport http oursprivacy https://ai.oursprivacy.com/mcp \
  --header "Authorization: Bearer YOUR_API_KEY"

Restart Claude Code, then ask: "List my running experiments." See the Claude Code MCP docs.

ChatGPT

1. Open Settings → Apps & Connectors → Advanced and turn on developer mode.
2. Under Apps, select Create.
3. Set the URL to https://ai.oursprivacy.com/mcp and add an Authorization: Bearer YOUR_API_KEY header.

Save, and the app is available in the message composer. See the OpenAI MCP docs.

Codex CLI

Add this to ~/.codex/config.toml, then restart Codex:

[mcp_servers.oursprivacy]
url = "https://ai.oursprivacy.com/mcp"
headers = { Authorization = "Bearer YOUR_API_KEY" }

See the Codex CLI docs.

Next Steps

• Manage Experiments with AI: run an experiment by prompt from start to finish.
• Platform API Authentication: choose an API key or OAuth for your AI tool.
• Platform OpenAPI Schema: the REST equivalents of everything the assistant can do.

Need Help?

Questions about setup, scopes, or which client to use? Reach out to support@oursprivacy.com.