More and more of our users don’t start their day in a dashboard. They start it in an AI assistant — writing code in Cursor or Claude Code, planning a client project in Claude, drafting an email in ChatGPT. So we brought Bannerproof there.
Bannerproof now has an MCP server. MCP — the Model Context Protocol — is the open standard AI assistants use to work with outside tools. Connect Bannerproof once, and your assistant can check websites for trackers that fire before consent, explain what it found and help you fix it, right in the conversation.
What it looks like
You: Check example-shop.example for cookie consent problems and tell me how to fix them.
Assistant: Score 38/100. Before any choice, Google Analytics and the Meta Pixel fire, and the TikTok Pixel keeps running after “Reject”. Set analytics_storage to “denied” by default in Consent Mode v2 and move both pixels to your consent tool’s Marketing category. Shall I re-check once that’s done?
You: Done — check again and give me a link for my client.
Assistant: Re-checked: 100/100, nothing fires before consent. Here’s a private report link for your client.
No copying results between tabs, no translating a report into tasks. And because the assistant explains things in your language, the findings make sense even if you’ve never opened the developer tools.
What your assistant can do
- Check any website — a free quick check, or a full audit with your audit credits. The assistant always asks before it spends a credit.
- Explain every finding — which trackers fired before consent or after “Reject”, the exact request as evidence, the reject button, Google Consent Mode v2 signals and what happens after “Accept”.
- Help you fix it — a concrete fix for each issue. In Cursor or Claude Code, where the assistant can see your code, it can make the change for you.
- Re-check right away — run a fresh scan after a fix and watch the score change.
- Monitor and report — add domains to weekly monitoring, see their history, and create a private report link for a client (white-label on the Agency plan).
It works with any consent tool — Cookiebot, Usercentrics, Borlabs, CookieYes, Complianz — and any website system, because Bannerproof checks the live site from the outside, like a first-time visitor. Fix instructions come in 13 languages.
How to connect
In claude.ai or ChatGPT: add a custom connector with the address https://api.bannerproof.app/mcp, sign in to Bannerproof and click “Allow”. That’s it — no token, no code.
In Claude Code, Cursor or VS Code: add the same address with your API token from Settings, or sign in when your editor asks.
Just want to try it? The free address https://api.bannerproof.app/mcp/free runs quick checks without an account.
Step-by-step guides are in our help center, and everything is on the MCP page.
What it costs
Nothing extra. The quick check is free. Full findings, fixes and report links come with any Bannerproof plan or with audit credits — no subscription needed for credits. You always pay on bannerproof.app; when something needs a plan or credits, your assistant gives you a direct link and you can continue right after paying.
Built to be safe
A connected assistant can read your scan results, run scans, add domains and create report links. It can’t see or change your billing, password or account settings, and it can’t buy anything. Sign-in uses OAuth, you can disconnect any assistant in Settings at any time, and scan results never include the content of the scanned websites — so a website can’t slip instructions to your assistant.
As always, Bannerproof is an automated technical check of what a site sends at the time of the scan, not legal advice.