What it does
Technical SEO Audit and Fix replaces the routine of downloading a crawl report and filing tickets by hand. You paste the crawl data — Core Web Vitals scores plus indexing errors, broken links, canonical conflicts, hreflang issues and schema.org problems — and it returns findings ordered by priority, a code-fix PR plan, and a checklist for verifying the fix afterwards. Each finding carries a traffic-impact estimate, which is what makes the ordering useful. A broken link and a failing LCP are not equally urgent, and the thing that decides which to do first is how much organic traffic each one is costing. That estimate is computed from your data, not asserted. The fix plan is the second half of the deliverable: a branch name, a title, and one item per non-ok finding naming the file and the change. Every plan is marked as requiring approval, and the agent has no path to opening or merging a pull request — it plans, a developer decides. For a marketing agent proposing changes to production code, that boundary is the right one. The verification checklist closes the loop, which most audits skip: re-run the crawl, confirm the vitals are back under threshold, request re-indexing for the previously excluded URLs, and confirm the PR was approved before merge. Each check is callable alone — `check_core_web_vitals`, `check_indexing`, `check_broken_links`, `check_canonical`, `check_hreflang`, `check_schema` — so you can ask one question without running the full audit. `plan_fix_pr` and `build_verification_checklist` stand alone too. Everything numeric is code-derived. The language model narrates the likely cause and the fix framing, and never invents a finding, a URL or a traffic number. No connectors and no credentials: the agent does no crawling of its own. If your host has Chrome DevTools or Search Console MCP connected they can supply the crawl data, and a GitHub MCP server can open the planned PR — but those are your host's connections. Built for SEO specialists and the frontend developers who end up doing the fixing.
Example prompts
- Here is our crawl export — what should we fix first?
- Which SEO issues are costing us the most organic traffic?
- Are our Core Web Vitals passing?
