Run Spronta from Claude, Codex, or your AI agent

Use the Spronta MCP to audit https://example.com. Call crawl_site with maxPages: 200, respectRobots: true, useSitemap: true and render: false. If it returns a jobId, poll crawl_status until it finishes and keep the reportId. Summarise the Health, GEO and accessibility scores, then call top_fixes. For the five highest-impact issue groups, use affected_urls and explain_issue. Return an executive summary, a prioritised action table with evidence and a 30-day plan. Separate measured facts from inference; do not dump raw JSON.

Use the Spronta MCP audit_url tool on https://example.com/pricing. Group the measured findings by severity. For each issue, use explain_issue where available and give me the evidence, why it matters, the safest fix and a concrete verification step. Do not invent site-wide impact from this one-page audit.

Use the Spronta MCP to crawl https://example.com and save the baseline reportId. Inspect top_fixes, affected_urls and explain_issue for the three highest-impact repeated findings. Trace those findings to shared templates or components in this repository and show me the proposed files, changes and risks before editing. Implement only evidence-backed, safe fixes; preserve unrelated work and run the relevant tests. After deployment, run a crawl with the same scope and use diff_reports against the baseline. Do not claim a live fix from the code diff alone.

Use the Spronta MCP to crawl https://example.com. Call top_fixes, then inspect every error-severity group with affected_urls and explain_issue. Trace repeated patterns to this repository, but stop for approval before changing canonicals, redirects, robots directives or noindex rules. Apply the approved fixes, test the build and tell me what must be deployed. After deployment, re-crawl with identical scope and use diff_reports to confirm which errors cleared and whether any regressions appeared.

Use the Spronta MCP to crawl https://example.com, wait for completion if needed, then call geo_gaps on the saved report. Break down the repeated gaps across indexable pages, inspect representative affected URLs and rank the work by coverage, confidence and effort. Prioritise changes that make content explicit, attributable and easy to extract. Distinguish technical findings from subjective editorial recommendations and do not promise AI citations.

Use list_reports to identify the latest completed report for https://example.com. Call top_fixes with category: geo and then geo_gaps. Give me only the GEO work, ranked by affected indexable pages, confidence and effort. For the top three items, include representative URLs and a one-line verification step.

Use Spronta audit_url on https://example.com/blog/my-post and focus on its GEO findings. Inspect the page source in this repository and propose evidence-backed improvements to answer clarity, headings, authorship, attribution and appropriate structured data. Show the plan before editing. Implement the approved changes and run the local build. Re-audit only after the change is deployed; do not promise inclusion or citation by an AI engine.

Use the Spronta MCP to crawl https://example.com, keep the baseline reportId and call geo_gaps. Use affected_urls for the repeated authorship and structured-data findings, then map those URLs to shared templates in this repository. Propose the smallest correct schema and byline changes before editing; never add schema for information that is not visible or true. After approval, implement and test. Once deployed, run the same crawl and use diff_reports to verify the gaps closed without regressions.

Use list_reports to find the latest completed baseline for https://example.com. Re-crawl the site with the same page limit, depth, rendering and include/exclude scope. Poll crawl_status if needed, then call diff_reports with the two report IDs. Tell me which issue groups were fixed, improved, worsened or introduced, plus any score and crawl-coverage changes. Give me a PASS, PARTIAL or FAIL verdict with evidence.

Use list_projects to find the Spronta project for https://example.com. Call project_trend for its score history and get_report for the newest completed report, then inspect top_fixes. Summarise the current Health and GEO scores, material changes since last week, new or worsening issue groups and the single highest-confidence action for this week. Keep it concise and link every claim to report evidence.

Use the Spronta MCP to re-crawl https://example.com with the same settings as its previous completed report, then call diff_reports. For every newly introduced error or high-impact warning, use affected_urls and explain_issue. Trace the regression to this repository and show me the suspected change and a minimal repair before editing. Implement approved fixes and run tests. Re-crawl only after deployment to confirm the regression is gone.

Use list_projects to check whether https://example.com already has a Spronta project. Do not create a duplicate. Propose a weekly schedule for Monday at 08:00 Europe/London with a 500-page limit and alerts for new errors and score regressions. Show the exact configuration and wait for my approval before calling create_project or set_schedule. After approval, apply it and report the resulting project ID and schedule.

Use list_reports to find the latest Spronta report for https://example.com, then call search_performance with both connected Google and Bing data and a limit of 100. Find high-impression pages with weak CTR, pages receiving demand that also have crawl findings and meaningful Google/Bing outliers. Inspect the relevant report evidence before recommending work. Cite the measured clicks and impressions returned by the tool; do not invent queries, rankings or conversions. Rank the backlog by impact, confidence and effort.

Use link_graph on the latest completed report for https://example.com. Identify important pages with low internal PageRank, pages deeper than three clicks, dead ends, orphan candidates and hub pages that could distribute authority better. Propose ten editorially relevant internal links. For each, give the source URL, destination URL, anchor concept and user benefit. Avoid repetitive exact-match anchors and sitewide footer links unless the evidence clearly supports them.

On the latest Spronta report for https://example.com, call search_performance and link_graph. Find pages with meaningful Google or Bing impressions but weak internal authority, excessive crawl depth or too few contextual inbound links. Return the ten best opportunities with the measured search data, graph evidence, recommended source pages and a safe anchor concept. Separate tool evidence from your inference.

Use the Spronta MCP explain_issue tool for the rule "geo-not-answerable". Explain what Spronta measured, why the signal matters, safe ways to improve it, common false positives and how I should verify a change. Do not imply that passing the rule guarantees an AI citation.

Use the Spronta MCP list_rules tool. Group the rule catalogue by category and severity, then give me a concise coverage map. Distinguish deterministic technical checks from findings that still require editorial or business judgement.