Model Context Protocol Architecture

Native MCP server for Claude, ChatGPT & Cursor

40 tools let your AI assistant audit sites, rank fixes, read site health and calibration, and propose changes. Every change it proposes waits for a person’s approval.

mcp://rankwyn — agent session
Paused
AI agent · connected via MCP
audit_site
{ site: "nordicroast.co" }
apply_metadata_fix
{ pages: 14, dry_run: true }
I’ve drafted 14 new descriptions. They’re waiting in your Approval Queue.
Human approval queue
empty
Waiting for agent actions…
/products/ethiopia-yirgacheffe+13 more
−meta_description: (empty)
+Single-origin Ethiopian Yirgacheffe, light roasted in Oslo. Bright citrus and jasmine notes. Ships in 24h.
Reject
Approve 14 changes
Published to WordPress
14 pages · 312ms · rollback available
Ask in plain English, from the AI client you already use.
0:14
https://app.rankwyn.com/mcp-execution-terminal
Claude Desktop Native
mcp: connected to api.rankwyn.com/mcp (protocol v1.2)
auth: master_token_verified (org: Rankwyn Operations)
ready for autonomous model tool invocation...
>

Standardized JSON-RPC Protocol

Exposed at api.rankwyn.com/mcp using official Model Context Protocol specifications for tool discovery and invocation.

Human Approval Queue

No AI assistant writes directly to WordPress or custom stacks without a team member reviewing and approving the diff in Rankwyn.

Multi-Site Context

The agent can inspect all managed properties linked to your organization in a single prompt context, comparing health scores and fixing shared issues.

MCP tools

40 tools for Claude, ChatGPT or Cursor. Agents can read everything below; every change they propose waits for a person’s approval.

Analyzer

14 tools
  • audit_url
  • get_audit_result
  • get_failed_checks
  • get_priority_fixes
  • explain_check
  • check_ai_bot_access
  • get_topical_gaps
  • get_link_opportunities
  • generate_fix
  • compare_snapshots
  • audit_site
  • get_site_health
  • get_crawl_insights
  • get_calibration

Sites and pages

10 tools
  • list_sites
  • test_connection
  • list_pages
  • get_seo_issues
  • get_page_seo_score
  • run_site_audit
  • get_page_schema
  • get_llms_txt
  • get_sitemap_entries
  • check_keyword_cannibalization

AI search and content

5 tools
  • get_ai_recommendations
  • get_aeo_tracking
  • score_entity_clarity
  • generate_content_brief
  • check_aeo_citation

Performance and analytics

3 tools
  • run_pagespeed_audit
  • get_search_console_summary
  • get_ga4_summary

Changes

8 tools

Proposed by the agent, applied only after a person approves.

  • apply_metadata
  • bulk_optimize
  • set_canonical
  • set_robots_meta
  • set_focus_keywords
  • upsert_schema
  • generate_llms_txt
  • record_aeo_citation

Developers and AI agents

REST and GraphQL API

A REST API with interactive docs, and a GraphQL endpoint for the same read surface.

MCP server

Claude, ChatGPT or Cursor can audit sites, rank fixes, read site health and calibration, and propose changes. Every change an AI makes needs a person’s approval.

AI assistance

You supply your own AI provider key. Keys are passed per call and never stored.

AI tools

An in-editor co-pilot, one-click field fixes, bulk optimize, bulk alt text, FAQ and How-to drafts, content briefs and a headline analyzer.

Watch: the problem and the fix

AI edits, with a human gate.

A 38-second film on why agents that write straight to production are risky, and how the approval queue makes them safe.

Read the transcript+
  1. IntroAI agents can fix websites in seconds. They can also break them in seconds.
  2. Beat 1Give an agent direct access, and one bad edit goes straight to production.
  3. Beat 2Rankwyn puts an approval gate between every AI agent and your client’s website.
  4. Beat 3Each change waits as a dry run preview, so you can see exactly what will change before it happens.
  5. Beat 4One click approves it. It publishes safely, with a full log and instant rollback.
  6. OutroRankwyn. AI powered fixes, with a human in control.

Ready to connect your AI assistant to your portfolio?

Set up your master token and start executing conversational audits in under 2 minutes.