Updated
What it is
An MCP server lets an AI assistant call tools on your behalf. DirectoryLens runs one at https://directorylens.com/mcp. Add that URL to Claude, ChatGPT, Claude Code, Cursor or any app that supports remote MCP servers, sign in once, and the assistant can read your Google Search Console data, plus Google Analytics and Bing Webmaster Tools if you linked them.
It is hosted, so there is nothing to install, no Google Cloud project to create and no API key to paste. It is read-only: the assistant can read reports, and it cannot change your site, your Search Console settings or anything in DirectoryLens except two things you ask for: starting a URL inspection run, and noting that you requested indexing for a URL.
Set it up in about two minutes
- Connect Search Console to DirectoryLens. Sign in with Google, pick your Search Console properties. DirectoryLens imports the 16 months Google keeps and syncs every day after that. The 5-day trial includes the MCP server.
- Add the server to your assistant. Use the URL
https://directorylens.com/mcp:- Claude (claude.ai, Desktop and mobile): Settings > Connectors > Add custom connector, paste the URL, then Connect. On Team and Enterprise plans an owner adds it first in Admin settings.
- ChatGPT: turn on developer mode in Settings (not every plan has it), create a connector with the URL and choose OAuth.
- Claude Code: run the command below, then
/mcpto sign in. - Cursor, VS Code and others: add a remote (HTTP) MCP server with the URL. Apps without OAuth sign-in can use a personal access token as an
Authorization: Bearerheader; you create one on your workspace page.
- Approve access on the DirectoryLens page that opens. The assistant then sees the same sites you do, nothing more. Disconnect it from your workspace page at any time; it loses access at once.
claude mcp add --transport http directorylens https://directorylens.com/mcpThe 31 tools
Every tool returns the same shape: the site, the date range, the data source, the units, notes on anything left out, and the data. Dates default to the 28 days ending on the last day Google has finalized, compared with the 28 days before, so the assistant does not compare a half-counted day with a full one.
| Group | Tools | What the assistant can do |
|---|---|---|
| Sites and performance | list_sites, portfolio_overview, site_overview, search_queries, search_pages, page_type_scorecard, query_type_breakdown | Clicks, impressions, CTR and position by site, search, page and page type, with the change against the previous period |
| Indexing | indexing_status, index_coverage_by_page_type, inspect_urls, url_index_history, request_indexing_queue, log_indexing_request | Share of each page type Google has indexed, per-URL inspections through Google’s URL Inspection API, and which pages to ask Google to index first |
| Growth | near_miss_keywords, new_queries, opportunities, ctr_benchmark, keyword_cannibalization, coverage_matrix, listing_gaps | Searches just off page one, new searches, pages clicked less than their position should earn, two pages competing, and (for directories) category × city demand without a page |
| Problems | content_decay, recovery_scorecard, thin_pages, risk_signals, url_migration_check, alerts | Traffic drops with the day they started, recovery after an update, pages Google never shows, quality signals and redirect checks after a URL change |
| Visitors and listings | engagement, bing_performance, listings, listing_report, category_city_rollups | Google Analytics contact clicks and AI-assistant referrals, Bing next to Google, and each listing’s searches, visits and contacts |
Directory-only tools (listings, coverage, rollups, risk signals, search intent) answer for sites set up as directories. Every other tool works on any website.
Hosted vs open-source Search Console MCP servers
There are good open-source Search Console MCP servers on GitHub. They are a thin layer over Google’s API: you create a Google Cloud project, enable the Search Console API, set up OAuth or a service account, run the server on your machine, and the assistant queries Google live. That is the right choice if you want full control and only need raw numbers.
DirectoryLens answers different questions, because it keeps and works on the data before the assistant asks:
| Open-source server | DirectoryLens | |
|---|---|---|
| Setup | Google Cloud project, OAuth client, local install | Paste one URL, sign in |
| History | What Google keeps: 16 months | 16 months on connect, then every day it stores, past Google’s limit |
| Page types | Not built in; the assistant sees URLs | Rules group URLs into listing, city, category and other page types |
| Indexing | Live URL Inspection calls, one URL at a time | Daily inspections spread over every page type, with estimates and alerts |
| Other sources | Search Console only | Search Console, Google Analytics 4 and Bing Webmaster Tools |
| Access | Your Google token on your machine | Read-only OAuth; your Google tokens never reach the assistant |
Privacy
The assistant reads reports through DirectoryLens with your access, and its provider receives what it reads under that provider’s own terms. DirectoryLens never gives the assistant your Google tokens, stores only hashes of the access tokens it issues, and you can disconnect any assistant from your workspace page. See the privacy policy.