MCP server

Endpoint

https://app.monoranks.com/api/mcp

Streamable HTTP, JSON responses, no session to keep. Send a key as a bearer header, or let the client sign you in with OAuth (below).

Connect

Claude Code

claude mcp add --transport http monoranks https://app.monoranks.com/api/mcp \
  --header "Authorization: Bearer mr_ws_…"

Claude.ai and Claude Desktop: Settings → Connectors → Add custom connector → the URL above. Add the key as a bearer header if the client asks for one; otherwise it signs you in.

ChatGPT and other OAuth clients: add https://app.monoranks.com/api/mcp as an MCP server. The client reads /.well-known/oauth-protected-resource, registers itself, and you sign in and approve the read scopes once. Write scopes are never granted this way.

Tools

Read

Tool	Does	Scope
list_sites	Websites the key covers, with ids	sites:read
portfolio	Every website in one call: scores and their change, issues, actions, last and next audit, and Search Console clicks (with search:read)	sites:read
site_summary	Scores, last audit, open issues, WordPress access, how fresh each connection is	sites:read
speed_results	Speed per tested page: lab test and real-visitor data	sites:read
audit_status	The latest audits, or one audit by audit_id, with pages crawled and the page budget	sites:read
list_issues	Issues, the same list as the Issues screen	issues:read
issue_detail	One issue with pages, evidence, explanation and suggested new values (100 at a time, suggestions_offset for more)	issues:read
list_actions	Prioritized actions with estimated gain and how long each has been open	issues:read
page_findings	Findings for one page, by id or address	pages:read
search_summary	Search Console totals, top pages and queries	search:read
search_rows	Search Console rows for a date range, by query, page, query and page, or day	search:read
ai_visibility	What AI assistants answered to your tracked questions, per week	ai:read
answer_gaps	Questions where AI answers name other websites but not yours	ai:read
geo_access	AI crawler access, llms.txt and entity signals from the last audit	ai:read
llms_txt	The live llms.txt and a ready-to-publish draft	ai:read
ai_crawler_rules	What robots.txt says to each AI crawler now	ai:read
agentic_results	Agentic browsing results with what to fix, or one re-run by job_id	ai:read

Act

Tool	Does	Scope
recheck	Queue a recheck of an issue	issues:recheck
run_agentic_check	Re-run the Agentic browsing check for one page	issues:recheck
start_audit	Start a full audit now, with an optional page budget	audits:run
apply	Approve an action’s new SEO titles, descriptions, canonicals or noindex and write them to WordPress	actions:apply
apply_llms_txt	Approve a new llms.txt and write it to WordPress	actions:apply
apply_ai_crawler_rules	Approve allow or deny rules for AI crawlers and write them into robots.txt	actions:apply

site_id is only needed when the key covers several websites; list_sites shows the ids. A missing scope comes back as a tool error that names the scope. The tools give the same answers as the REST API, so the details there apply here too.

Example conversation

“Which of my websites lost the most clicks this month, and what is the top action for it?”

The assistant calls portfolio once, picks the website with the biggest drop, then calls list_actions for it, and answers with the action and its estimated gain. If you say “apply it”, it reads the suggested values with issue_detail, calls apply, and the change shows up in the app’s activity with Undo.