A practical MCP workflow

How to run an SEO audit with an AI agent and MCP

Connect ToolCargo Site Audit to your MCP client, audit your important public URLs, and ask your agent to turn the returned evidence into a prioritized fix list. Re-run the same pages after changes to check what improved.

Built for: Agencies and developers preparing a website launch or investigating an indexing regression.

What to connect

Create a ToolCargo account and use OAuth or an API key with a supported MCP client. Hosted connectors share your plan’s call quota; connect each required MCP endpoint separately. Review provider permissions before starting.

Client setup and test status · Current plans and limits

Run the workflow

1. Choose the pages that matter
Start with the homepage, your main product page and a page that should bring enquiries. Use their public production URLs. A staging page can have intentionally different indexing settings.
2. Run the audit and keep its evidence
Use audit_page for one URL or audit_site for a capped selection. Review status codes, canonical tags, robots directives, titles, descriptions and social preview findings. Keep the dated report so a future comparison has a baseline.
3. Separate blockers from improvements
An accidental noindex directive or a broken response deserves attention before a title rewrite. Confirm that each reported canonical is the intended public URL. A score helps sort checks; it is not a Google ranking signal.
4. Fix, re-run and compare
Deploy the fixes, audit the same URLs again and compare the saved results. Use Google Search Console to investigate crawling and indexing separately; a passing HTML audit does not establish Google index coverage.

A prompt to try

Audit our homepage, pricing page and main service page. For each issue include the URL, observed evidence, why it matters and one proposed fix. Put indexing blockers first. After we deploy, re-run those URLs and compare with the saved reports.

What a useful result looks like

Example handover: URL → observed noindex directive → remove it from the production page → re-run the audit → inspect Google indexing separately. Preserve the actual observed value rather than inventing a score or predicting traffic.

Know the limits

Site Audit reads served HTML and does not execute JavaScript. The site audit has a page cap and is not a complete crawl. Accessibility and asset checks are static checks, not a full accessibility assessment or a Core Web Vitals measurement.

Common questions

Can I use this SEO audit in Claude?

Site Audit has recorded end-to-end tests in the clients marked tested on its connector page. Follow the Claude or Claude Code setup guide for your client. Other client setups are documented with their test status.

Does an audit fix my website automatically?

No. It returns findings and suggested fixes. Your agent or developer must review and apply changes in your website project, then verify the deployed pages.

Do I need a paid plan for the first audit?

ToolCargo has a shared free plan with call limits. The browser demo on the Site Audit page also has rate limits. Check the pricing page for current quotas before auditing many pages.

References and tool documentation

Use the provider’s documentation to check the underlying concepts, and ToolCargo’s references for the exact tools, inputs and limits.

Google: robots meta tags and X-Robots-Tag
Check how page-level directives affect indexing and why crawlers must be able to read them.
Google: choosing a canonical URL
Understand canonical signals before changing a page’s preferred URL.

Tool references for this workflow

Continue with the tools

View the main connector Set up your client