Connect with MCP
Let Claude, ChatGPT or Gemini read your audit results through MCP.
MCP (Model Context Protocol) is an open standard that lets an AI assistant call tools on a server. Our server gives the assistant read access to your audit results, so you can ask things like "which pages have no title?" or "what changed since the last scan?" and get answers from your own data.
Server address
https://search-scanner.ucartz.com/mcpIt uses the Streamable HTTP transport and answers with plain JSON. Your assistant needs to reach it over the internet, so local-only assistants must run it through a client that supports remote servers.
How connecting works
- You add the address above in your assistant. It discovers our sign-in service by itself.
- It opens a Ucartz consent page in your browser. You sign in (if needed) and see which app asks for what.
- You allow or deny. The assistant gets a short-lived token (1 hour) and a refresh token (30 days). It never sees your password or Google account.
- Under Settings, Connect AI you can see every connected app, when it was last used, and disconnect it.
Most assistants register themselves (Dynamic Client Registration), so there is nothing to copy. If yours asks for a client ID and secret, create them in Settings, Connect AI. The secret is shown once and stored only as a hash.
Permissions
seo:read: read your sites, scans, pages, issues and reports.seo:write: also start, stop and resume scans. Limits and the one-scan-per-site rule are the same as in the app.
You choose on the consent page. A token with only seo:read does not even list the scan-control tools.
Security
- OAuth 2.1 with mandatory PKCE (S256) and exact redirect address matching.
- Tokens are random 256-bit values, bound to your account, the app and this server, and only their hash is stored.
- Every tool call runs as you, with the same ownership checks as the website. Other accounts' data is unreachable.
- Refresh tokens rotate on every use. Disconnecting an app deletes its tokens at once.
Tools
This table is generated from the server's own tool list.
| Tool | What it does | Scope |
|---|---|---|
| list_sites | List the sites in this account with id, domain, latest health score, page counts, issue counts and the latest scan. Start here to get site ids. | seo:read |
| get_site_overview | Health score, issue counts by severity, top issues, score trend and scan status for one site, or for all sites when site_id is left out. | seo:read |
| start_scan | Queue a new scan of a site you own. Fails with a conflict if a scan is already queued or running for that site. Uses the same limits as the web app. | seo:write |
| get_scan_status | Latest scan of a site: status, pages done and total, rate, time left, score when finished. | seo:read |
| stop_scan | Ask a queued or running scan to stop. A running scan finishes its current slice first. It can be resumed later. | seo:write |
| resume_scan | Continue a stopped scan where it left off. Fails if another scan is active for the site. | seo:write |
| list_pages | Pages known for a site with status code, title, and a problem count per area. Filter by state or search the address. | seo:read |
| get_page_report | Everything known about one page: fetch facts, title, description, canonical, headings, social tags, score, all problems with details, links summary and PageSpeed results. Use a page id from list_pages or list_issues. | seo:read |
| list_issues | Problems found, with severity (critical, medium, low), code, page and details; includes totals by severity and by code. | seo:read |
| list_audit_area | One audit area (title, description, headings, canonical, social, url, schema, status, links, sitemap, content, assets, performance): pages checked, how many are fine or have problems, and the pages with their problems. | seo:read |
| list_links | Internal or external links found on a site's pages. group=target gives one row per target page, most linked first. External links are listed only, never checked. | seo:read |
| list_assets | Images, scripts, stylesheets and share images used by a site, with size, status, compression and the number of pages using each. | seo:read |
| get_content_summary | Word counts per page: how many pages are empty, thin or long, plus the shortest pages and text-to-HTML ratio. | seo:read |
| get_performance | Google PageSpeed (Lighthouse) scores and Core Web Vitals for the tested pages of a site, worst first. Empty when PageSpeed is not set up by the administrator. | seo:read |
| get_schema_summary | Structured data (schema.org) types found on a site and how many pages use each. | seo:read |
| list_sitemaps | The sitemaps added for a site with their status and URL count. | seo:read |
Step by step for each assistant
- Claude (claude.ai, Desktop, Claude Code)
- ChatGPT
- Gemini (Gemini CLI)
- Cursor and VS Code
Problems? See Troubleshooting.