Claude can inspect consent code and test browsers. International checks need regional network routing, fresh sessions for each consent choice, and saved evidence you can compare after a fix.
CookieTrail supplies those checks through its dashboard, application programming interface (API), and Model Context Protocol (MCP) connection. MCP lets Claude call audit tools and retrieve evidence while it uses your repository to investigate the cause.
Create an international consent test plan
Select visitor locations and choose whether to include a retest. Each audit covers all three consent paths. Copy the Claude prompt or download a QA worksheet for the selected scope.
Replace the example URL with your public page. This builder does not visit or audit the URL.
3 locations × 3 consent paths
That is 3 audits, each covering first load, accept, and reject. Actual coverage can be incomplete. US routing options do not establish a specific state's location.
Connect CookieTrail before using the prompt.
Read the generated prompt
Use the connected CookieTrail MCP tools to audit https://example.com. Run a separate audit for each of these visitor locations: Germany (de), United Kingdom (uk), US East (us-east). Check workspace access and remaining usage before starting. Each audit measures first load, accept all, and reject all in fresh browser sessions. Retrieve and save the audit IDs, report links, location verification, consent-state evidence, cookies, measured requests, and coverage limitations. Keep incomplete tests inconclusive. Inspect my actual repository or tag configuration before confirming a cause or changing code. Distinguish measured findings from proposed explanations. Summarize the measured findings and the extra evidence needed to investigate them. Propose a fix and a matching regional retest plan. Do not treat a request alone as proof of a consent violation or this bounded check as legal certification.
Start with the implementation
Ask Claude to trace consent defaults, tracking scripts, and accept/reject callbacks. Repository access helps it locate hardcoded scripts and Google Tag Manager (GTM) integrations; browser access lets it inspect the running page.
Claude Code's Chrome integration uses your connected browser. Its checks inherit that browser's network location and session state unless you change them. Source inspection can identify a likely cause, but can't establish what an untested country received.
- Claude Code with Chrome Anthropic's browser testing capabilities and setup.
Define the international test plan
Specify the URL, locations, and expected behavior. Two locations with three paths mean six observations; another page with different scripts needs its own checks. Regional differences can be intentional, so record what each configuration should do.
- Verify the requested country using regional network routing.
- Start fresh for first load, acceptance, and rejection.
- Save the stored choice, cookies, measured requests, screenshots, and coverage gaps.
- After deployment, repeat the URL, location, and path that produced the finding.
Connect Claude to regional checks
You can build a repeatable regional browser harness with a coding agent. Your team then maintains routing, session isolation, banner interactions, and evidence capture as sites and consent management platforms (CMPs) change. CookieTrail manages that capture workflow so Claude can work from saved evidence.
For CookieTrail's managed workflow, choose Create key & connect Claude in your workspace and copy the setup message into Claude. The hosted MCP connection can start audits, retrieve evidence, inspect saved findings, and verify them after a change.
- CookieTrail API and MCP setup Connection instructions and supported operations.
Investigate the saved finding
Have Claude read the evidence before changing code. A known cookie present after verified rejection is an observation; a tag bypassing the consent callback is a hypothesis until Claude inspects the source or configuration.
CookieTrail's request capture is bounded and excludes bodies, headers, and query values. Requests alone don't establish their payload or a consent violation, and browser checks don't inspect server-side tracking. Ask Claude what extra evidence it needs.
After deployment, rerun at the original URL and location. Use the comparison and finding history to see what resolved, remains, or couldn't be verified.
Keep the plan with the release checklist
Repeat the plan when you change the CMP, publish a GTM container, or add a script. Use recurring monitoring where it fits. Keep the finding, owner, confirmed cause, and retest together so another developer can continue without reconstructing a chat.