Rankealo

SEO agents

Rankealo exposes agents in three layers: conversational SEO inside the app, programmatic tools via API/MCP, and an optional GitHub App that opens code fix PRs on your repo.

Canonical product page

For positioning, demos, and install CTAs, see the marketing page /for-agents. This doc is the reference map for developers.

ChatSEO agent (in-app)#

The SEO agent answers questions about your domain using live GSC, keyword pool, visibility, and content data — not generic SEO advice. Access it from the app dashboard or via POST https://app.rankealo.ai/api/v1/seo-agent/chat with a write-scoped API key.

MCP equivalent: chat_seo_agent on the MCP server.

MCP & REST API#

For Claude, Cursor, Windsurf, or Codex, install @rankealo/mcp-server with a domain-scoped key. Read tools fetch calendar, articles, GSC, and visibility; write tools publish, reschedule, and change settings (Growth + write scope).

Same data and actions are available over REST at https://app.rankealo.ai/api/v1/*. For Zapier or Make, use Webhooks + OpenAPI import against the same surface.

GitHub coding agent#

The coding agent watches SEO/GEO health findings and opens pull requests on your connected GitHub repo — robots.txt, meta tags, schema, canonical fixes, and similar technical items. It does notreplace CMS publishing; it patches your site's source code.

Setup, permissions, and autopilot behavior are documented on /for-agents and in the in-app agent workspace (/agents/setup when signed in).

CI audit check (coding agent)#

Gate deploys on SEO/GEO health, like a test suite. POST /api/coding-agent/ci-check runs a bounded single-page audit (page + robots.txt + llms.txt, same methodology as the full GEO audit — no crawl) against any URL — a preview deploy or production — and returns passed: true|false plus severity-ranked findings (critical | high | medium | low | info). fail_on sets the threshold; the endpoint always answers 200 and your pipeline decides the exit code.

Authenticate with your domain API key (rnk_…, read scope is enough). Copy-paste GitHub Actions step:

yaml
- name: Rankealo SEO CI check
  env:
    RANKEALO_API_KEY: ${{ secrets.RANKEALO_API_KEY }}
    TARGET_URL: ${{ env.PREVIEW_URL }} # or your production URL
  run: |
    RESULT=$(curl -sf -X POST "https://app.rankealo.ai/api/coding-agent/ci-check" \
      -H "Authorization: Bearer $RANKEALO_API_KEY" \
      -H "Content-Type: application/json" \
      -d "{\"url\": \"$TARGET_URL\", \"fail_on\": \"high\", \"pr_number\": ${{ github.event.pull_request.number || 'null' }}}")
    echo "$RESULT" | jq '{passed, failOnThreshold, counts}'
    [ "$(echo "$RESULT" | jq -r .passed)" = "true" ] || exit 1

Add ?format=junit (or "format": "junit" in the body) for JUnit XML that any CI test-report ingester understands. When you pass pr_number (and/or head_sha) and the domain has the GitHub App connected, the result is also posted to the PR — as a GitHub check run when the App has checks:write, otherwise as a PR comment with the findings table.

Feature flag

The endpoint ships behind CODING_AGENT_CI_CHECK_ENABLED (default off) and returns 404 until enabled. Preview deploys that send noindex will surface that as a finding — audit the production URL, or treat it as expected on previews.

Which agent when#

  • Content & strategy questions → ChatSEO / MCP read tools.
  • Publish or reschedule from an external harness → MCP/API write tools.
  • Fix technical SEO in your codebase → GitHub coding agent.

Still stuck? Contact support

Reading is step one. Measuring your AI visibility is step two.

Rankealo tracks how often your brand is mentioned and cited across the major AI engines, then helps you publish the pages that close the gaps.