DataForSEO OnPage API

Provenance

This is an experimental project to test how OpenCode, plugged into frontier LLMs (OpenAI GPT-5.2), can help generate high-fidelity agent skill files for API integrations.

When to Apply

• "crawl a site", "technical SEO audit", "find broken links"
• "duplicate content", "duplicate tags", "non-indexable pages"
• "redirect chains", "waterfall performance", "resource issues"
• "run Lighthouse at scale", "performance audit"

Integration Contract (Language-Agnostic)

See references/REFERENCE.md for the shared DataForSEO integration contract (auth, status handling, task lifecycle, sandbox, and .ai responses).

Task-based Crawl Lifecycle

• Create a crawl via task_post.
• Wait for completion by polling tasks_ready (or use webhooks if supported in the endpoint docs).
• Fetch results via summary, pages, resources, and other specialized result endpoints.
• To stop an in-progress crawl, call force_stop.

Steps

1. Identify the exact endpoint(s) in the official docs for this use case.
2. Choose execution mode:

- Live (single request) for interactive queries - Task-based (post + poll/webhook) for scheduled or high-volume jobs

1. Build the HTTP request:

- Base URL: https://api.dataforseo.com/ - Auth: HTTP Basic (Authorization: Basic base64(login:password)) from https://docs.dataforseo.com/v3/auth/ - JSON body exactly as specified in the endpoint docs

1. Execute and validate the response:

- Check top-level statuscode and each tasks[] item status - Treat any statuscode != 20000 as a failure; surface status_message

1. For task-based endpoints:

- Store tasks[].id - Poll tasksready then fetch results with taskget (or use postbackurl/pingbackurl if supported)

1. Return results:

- Provide a normalized summary for the user - Include the raw response payload for debugging

Inputs Checklist

• Credentials: DataForSEO API login + password (HTTP Basic Auth)
• Target: keyword(s) / domain(s) / URL(s) / query string (depends on endpoint)
• Targeting (if applicable): location + language, device, depth/limit
• Time window (if applicable): date range, trend period, historical flags
• Output preference: regular vs advanced vs html (if the endpoint supports it)

Example (cURL)

curl -u "${DATAFORSEO_LOGIN}:${DATAFORSEO_PASSWORD}"   -H "Content-Type: application/json"   -X POST "https://api.dataforseo.com/v3/<group>/<path>/live"   -d '[
    {
      "<param>": "<value>"
    }
  ]'

Notes:

• Replace <group>/<path> with the exact endpoint path from the official docs.
• For task-based flows, use the corresponding taskpost, tasksready, and task_get endpoints.

Docs Map (Official)

• Overview: https://docs.dataforseo.com/v3/on_page/overview/
• Task POST: https://docs.dataforseo.com/v3/onpage/taskpost/
• Tasks Ready: https://docs.dataforseo.com/v3/onpage/tasksready/
• Summary: https://docs.dataforseo.com/v3/on_page/summary/
• Pages: https://docs.dataforseo.com/v3/on_page/pages/
• Resources: https://docs.dataforseo.com/v3/on_page/resources/
• Force Stop: https://docs.dataforseo.com/v3/onpage/forcestop/

Lighthouse:

• Overview: https://docs.dataforseo.com/v3/on_page/lighthouse/overview/

Business & Product Use Cases

• Build a technical SEO crawler product (issues, prioritization, exports).
• Monitor site health after deployments (regressions: redirects, noindex, broken links).
• Create an automated QA gate for SEO before release.
• Performance improvements: integrate Lighthouse outputs into engineering backlogs.
• Agency deliverables: repeatable site audits with remediation tracking.
• Template validation: verify titles/meta/canonicals across large sites.

Examples (User Prompts)

• "If you don't have the skill installed, install dataforseo-onpage-api and then continue."
• "Install the OnPage skill and crawl our site to find broken links, redirect chains, and non-indexable pages."
• "Run a technical SEO audit and produce a prioritized fix list."
• "Check for duplicate titles/meta across the site and summarize affected URLs."
• "Generate a report of slow pages with waterfall diagnostics and top heavy resources."
• "Run Lighthouse checks for these 50 URLs and highlight regressions."