Build on AgenticSEO

MCP server — point an MCP-capable client (Claude Desktop, Cursor, an OpenAI-compatible agent runtime, your own orchestrator) at the server URL and it discovers the tool list, schemas, and descriptions automatically.

{
  "mcpServers": {
    "agenticseo": {
      "url": "https://hbzbbsodddymrhmreqkd.supabase.co/functions/v1/mcp",
      "headers": { "Authorization": "Bearer $AGENTICSEO_TOKEN" }
    }
  }
}

REST API — the same functions as ordinary endpoints for backends and scripts that do not speak MCP.

MCP  https://hbzbbsodddymrhmreqkd.supabase.co/functions/v1/mcp
REST https://hbzbbsodddymrhmreqkd.supabase.co/functions/v1/api

Each function below is one MCP tool and one REST endpoint. Parameter schemas, response shapes, examples, and per-call credit metering are documented in the full integrator reference.

Authentication is a bearer token on every request. The typical audit flow is discover → analyze → poll → read.

# 1. Enumerate the pages on a site
curl -X POST "https://hbzbbsodddymrhmreqkd.supabase.co/functions/v1/api/v1/discovery" \
  -H "Authorization: Bearer $AGENTICSEO_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"url":"https://example.com"}'

# 2. Queue an analysis over the pages you care about
curl -X POST "https://hbzbbsodddymrhmreqkd.supabase.co/functions/v1/api/v1/reports" \
  -H "Authorization: Bearer $AGENTICSEO_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"url":"https://example.com","pages":["/","/pricing","/product"]}'

# 3. Poll the job, then read the finished report
curl "https://hbzbbsodddymrhmreqkd.supabase.co/functions/v1/api/v1/jobs/JOB_ID" -H "Authorization: Bearer $AGENTICSEO_TOKEN"
curl "https://hbzbbsodddymrhmreqkd.supabase.co/functions/v1/api/v1/reports/REPORT_ID" -H "Authorization: Bearer $AGENTICSEO_TOKEN"

Tokens are least-privilege: read-only by default, with write and spend capability granted per scope.

• read — All GET endpoints (list_reports, get_report, get_visibility_scores, list_prompts, get_prompt_history, get_ai_crawler_access, list_publish_targets, get_llms_txt, get_gsc_performance, list_gsc_properties, list_products, get_product_visibility).
• analyze — Write / enqueue endpoints (analyze_site, analyze_ai_visibility, enable_visibility_monitoring, sync_product_catalog, scan_product_visibility).
• faq:draft — draft_faqs_for_gaps · POST /v1/faqs/draft.
• faq:publish — publish_faq_to_target · POST /v1/faqs/publish · resolve_gaps · POST /v1/gaps/resolve.
• competitors:write — add_competitors · remove_competitors · discover_competitors (persist=true). GET list_competitors and preview discover_competitors do NOT require this scope.
• gap_pages:read — list_gap_pages · get_gap_page (MCP) · GET /v1/gap-pages · GET /v1/gap-pages/{id_or_slug}.
• gap_pages:publish — publish_gap_pages · unpublish_gap_page (MCP) · POST /v1/gap-pages/publish · POST /v1/gap-pages/unpublish.
• work:write — record_went_live_date · send_recommendation_to_opportunities · POST /v1/work/went-live · POST /v1/visibility/recommendations/send.
• connect_links:send — send_client_connect_link · POST /v1/connect-links.
• pages:publish — publish_new_page · unpublish_new_page · POST /v1/pages/publish · POST /v1/pages/unpublish.
• commerce (account entitlement, not a token scope) — Headless product / SKU visibility: sync_product_catalog · scan_product_visibility · list_products · get_product_visibility, over both transports.

Tokens are issued by Traction Gap Partners; there is no self-serve key page. Email support@tractiongappartners.com with your company, the websites or brands the token should see, whether you need MCP, REST, or both, the scopes you need, and the technical owner of the integration. Tokens are scoped to your account and its clients, rate-limited per hour, and every call is attributed in the audit log. Headless commerce (product/SKU) functions require an Enterprise or OEM entitlement.

Programmatic access is available on Enterprise and OEM agreements. See pricing or write to info@tractiongappartners.com.

The public /docs/quickstart page needs no sign-in. It covers both endpoints, the rule that the key stays on the server, every supported function with its permission, the audit sequence, and a paste-ready brief for an AI coding engine. A Word version downloads from the page.

The complete integrator reference — parameter schemas, response bodies, MCP system prompt, calling flows, error envelope, rate limits, credit metering, and a downloadable Word version — lives at /docs/api. It is behind sign-in because it covers token handling and account-specific detail; ask support for access alongside your token request.

• AI agents for SEO — how autonomous agents run the audit-and-fix loop.
• Free llms.txt generator — no token required.
• Security — tenant isolation, secret handling, publishing safety.