dotTide
OfficialRetrieves live public statistics for Firefox Add-ons, enabling agents to check extension store numbers alongside other site and product metrics.
Retrieves public GitHub repository statistics such as stars, with support for a read-only token to avoid anonymous rate limits.
Reads GA4 traffic data across multiple Google accounts and properties, reporting sessions by source and medium and separately counting referrals from AI assistants such as ChatGPT, Perplexity, Copilot, Gemini, and Claude.
Reads Search Console performance data across multiple Google accounts and properties, including clicks, impressions, CTR, and average position by query or page for arbitrary date ranges, with period-over-period comparisons.
Retrieves live public statistics for WordPress.org plugins, such as active installs and other store metrics, without requiring an API key.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@dotTide compare last week's search clicks to the week before"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
dotTide
dotTide is a read-only MCP server and command-line tool that gives Claude, Cursor, Codex and other AI agents the numbers for every site you run. It reads Google Search Console and Google Analytics 4 (GA4) across multiple Google accounts, plus WordPress.org, Firefox Add-ons and GitHub stats, and writes a weekly snapshot for scripts and agents.
Nothing gets added to your sites. dotTide only reads numbers you already have, and your keys stay on your computer.
Why
If your sites sit under different Google accounts, checking them means logging in and out of each one. The weekly check slips to monthly, and any AI agent you ask works from stale numbers. Store numbers for plugins, extensions and repos live in yet more places.
dotTide reads all of it with one read-only key, and answers the same way in Claude, on the command line, and in a weekly JSON file.
Related MCP server: SEOAnalytics Read MCP
What it does
sites: every configured site, which key covers it, and whether Search Console, GA4 and Bing access works when you ask.search_queries: clicks, impressions, CTR and average position by query or page, for any date range, with an optional comparison to the previous period and filters for country, device and search type.traffic: GA4 sessions by source and medium, with AI assistant referrals (ChatGPT, Perplexity, Copilot, Gemini, Claude and others) counted separately, with the landing pages each assistant sent visits to (landing_pages). Only for sites that have GA4.store_stats: live public numbers that need no key: WordPress.org plugins, Firefox Add-ons, GitHub stars.opportunities: what to work on next: queries in striking distance, pages and queries losing clicks, pages and queries slipping in position (slipping_pages,slipping_queries), and page-one queries with a low click-through rate for their position. When that last list is empty because too few queries rank on page one,low_ctr_notesays so and gives each position band's count. List your brand names asbrand_termsin the profile and searches for them stay out of the low click-through check, since they get far more clicks at the top.page_queries: the queries that bring people to one page, and how much of the page's traffic Google hides behind anonymized queries.competing_pages: queries where two or more of your pages split the impressions, biggest overlaps first.index_health: every sitemap's state, and whether the top pages are in Google's index, with a plain list of problems to fix.core_web_vitals: real-user LCP, INP and CLS with Google's ratings when there's enough data, plus a lab run, from PageSpeed Insights.bing_queries: Bing Webmaster Tools clicks, impressions and position by query. Bing feeds Copilot.discover: finds every Search Console property and GA4 property the key can read, and writes the profile entries for you, with abrand_termssuggestion from each domain name, commented out until you check it.Ready-made prompts: a weekly review, what to work on, and an index check, shown as commands in Claude and other MCP clients.
Weekly snapshot: last full week against the week before, one JSON file per profile, for scripts and agents that read files.
How access works
No site moves to a new Google account. You create one Google service account, and whichever Google account owns a site invites that service account's email:
Search Console: as a Restricted user.
GA4: as a Viewer.
The service account can only read, and you can remove it from any site in one click. Bing is optional and uses one API key from Bing Webmaster Tools that covers all your sites there. Its key lives in your system's secret store: the macOS keychain, Windows Credential Manager or the Linux Secret Service.
Quick start
You need macOS, Windows or Linux, and uv. There is nothing to clone: uvx downloads dotTide the first time you run it.
Write a starter profile, then add a plugin slug, an add-on slug or a GitHub user under
[stores]in it:
uvx dottide initCheck that it runs. Store numbers need no Google setup at all:
uvx dottide store_statsSet up Google access by following Google setup. It takes about ten minutes, plus a minute per site.
Store the key, let dotTide find your sites, and check access:
uvx dottide import-key dottide-main ~/Downloads/<downloaded-key>.json
uvx dottide discover --write
uvx dottide sitesOptional: add Bing numbers by following Bing setup. It takes about five minutes.
Connect it to Claude Code:
claude mcp add --scope user dottide -- uvx dottideTo update later, run uvx dottide@latest sites once; after that uvx dottide uses the new version.
Documentation
Google setup: the service account, click by click, and sharing each site.
Bing setup: optional Bing Webmaster Tools numbers with one API key.
Configuration: profiles, every field, keys and environment variables.
Using dotTide: each tool and its options, MCP clients, the weekly snapshot and its format, and fixes for common errors.
Asking an AI agent for your numbers: example prompts for weekly checks, content to refresh, AI referrals and store numbers, plus a saved Claude Code command.
Connecting AI clients: Claude Code, Claude Desktop, Cursor, Windsurf, VS Code, Codex, Gemini CLI, Grok, and agents without MCP.
Good to know
Search Console data settles about three days after the fact, so default ranges end three days ago.
Every Google scope dotTide asks for is read-only, and no tool can change anything.
GitHub allows 60 anonymous requests an hour per IP. Set
DOTTIDE_GITHUB_TOKENto a read-only token if you hit the limit.
Contributing
See CONTRIBUTING.md. Security reports go through SECURITY.md.
License
The code is released under the MIT License. The dotTide name and icon are not covered by the license, so please use your own name and icon if you ship a copy.
Available Tools
11 toolsbing_queriesBRead-only
Bing Webmaster Tools clicks, impressions and average position by query, plus site totals, for a range. Bing feeds Copilot. Needs a [bing] key in the site's profile. Bing updates these numbers about once a week.
Inputs: site (required), end, days, limit, +1 more.
competing_pagesARead-only
Queries where two or more pages of the site each get at least min_impressions, so they
split the clicks (cannibalization). For each query: every page with its clicks, impressions
and position, the page with most clicks and its share. Biggest overlaps first. Same dates
and filters as search_queries.
Inputs: site (required), end, days, limit, +5 more.
core_web_vitalsARead-only
Core Web Vitals for a page (default: the site's home page) from PageSpeed Insights. page_field and site_field hold real-user 75th-percentile LCP, INP and CLS with Google's good / needs improvement / poor rating, when Google has enough visits; lab holds one Lighthouse run (performance score, LCP, CLS, TBT, FCP, Speed Index). strategy is "mobile" or "desktop". One run takes 10 to 30 seconds.
Inputs: site (required), page, strategy.
discoverBRead-only
Every Search Console property and GA4 web stream the key can read, matched by domain,
with ready-to-paste [[sites]] blocks for any site not in a profile yet. Each block has a
commented-out brand_terms suggestion from the domain name. Reads only; the user adds the
blocks with dottide discover --write.
Inputs: profile.
index_healthARead-only
Is the site healthy in Google's index? Every sitemap Search Console knows (errors,
warnings, when Google last read it, URLs submitted) and the index status of the top pages
pages by clicks over the last days days (verdict, coverage, last crawl, Google's chosen
canonical). problems lists anything to fix in plain words. Reads only: it never submits a
sitemap or requests indexing. Each page checked uses one URL Inspection call.
Inputs: site (required), days, pages.
opportunitiesARead-only
What to work on next for a site, from Search Console: queries in striking distance (position 8 to 20, by impressions), pages and queries that lost the most clicks against the previous period, pages and queries whose average position dropped the most (slipping_pages, slipping_queries, with min_impressions in both periods), and page-one queries whose CTR is under half the site's own median for that position. When low_ctr is empty because no position band has three queries, low_ctr_note says so and low_ctr_band_counts gives each band's count. Every item carries its numbers. Same date rules and filters as search_queries, except Discover and Google News, which have no query rows.
Inputs: site (required), end, days, limit, +5 more.
page_queriesARead-only
The queries that bring people to one page (full URL), with the page's own totals and
hidden: the clicks and impressions Search Console counts for the page but leaves out of
its query rows, because Google anonymizes rare queries. Same dates, compare and filters as
search_queries.
Inputs: page (required), site (required), end, days, +6 more.
search_queriesARead-only
Search Console clicks, impressions, CTR (percent) and average position, by "query" or
"page". Dates are YYYY-MM-DD; without them the range is the last days days ending three
days ago. compare=true adds the previous period of the same length and the change.
Optional filters: country (three-letter ISO code like "usa"), device ("desktop", "mobile"
or "tablet") and search_type ("web", "image", "video", "news", "discover" or
"googleNews"; default "web"). Discover and Google News have page rows only. The result
echoes the filters it used.
Inputs: site (required), by, end, days, +6 more.
sitesARead-only
Every configured site, which key covers it, and whether Search Console and GA4 access works at the moment of the call.
Inputs: no named parameters.
store_statsCRead-only
Live public store numbers that need no keys: WordPress.org plugins (installs, rating, downloads last 7 days), Firefox Add-ons (daily users, ratings), GitHub stars.
Inputs: profile.
trafficBRead-only
GA4 sessions by source and medium, plus AI assistant referrals (ChatGPT, Perplexity, Copilot, Gemini, Claude and others) totalled separately. Each assistant lists the landing_pages its visits started on, most sessions first. Only for sites with GA4.
Inputs: site (required), end, days, limit, +2 more.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 11 tools
Most tools target distinct resources, but search_queries (which accepts a 'page' dimension) and page_queries overlap significantly—an agent could reasonably reach for either when asked for a page's queries. bing_queries vs search_queries is clearer due to the source distinction, and the analytic tools (opportunities, competing_pages, index_health, core_web_vitals) are well separated.
Nearly all names are lowercase snake_case noun phrases (bing_queries, search_queries, core_web_vitals, index_health, competing_pages). Minor deviation: several single-word nouns (sites, traffic, opportunities) and one verb (discover) break the compound pattern slightly, but the set remains readable and predictable.
Eleven tools sit squarely in the well-scoped range, and each covers a distinct slice (per-source queries, GA4 traffic, discovery, store stats, opportunity analysis, cannibalization, index health, CWV). No obvious padding or redundancy.
The surface covers the core read-only SEO analytics lifecycle: discovery, traffic sources, keyword/query performance across GSC and Bing, index/sitemap health, and Core Web Vitals, plus derived opportunity analysis. Gaps exist (no backlink analysis, no sitemap submission—explicitly read-only by design), but these are minor workarounds.
Maintenance
Related MCP Connectors
Read-only Google Search Console MCP server for AI assistants.
Google Search Console, GA4 and PageSpeed MCP server for Claude, ChatGPT and Cursor. Free, hosted.
Read-only Google Analytics 4 MCP server for AI assistants.
Hosted MCP server for GA4, Google Ads and Search Console. Google OAuth, nothing to install.
Related MCP Servers
- AlicenseAqualityDmaintenanceMCP server for querying Google Search Console data — search analytics, URL inspection, sitemap monitoring, and more — read-only tools for any MCP-compatible AI client.733 PyPIApache 2.0
- FlicenseNot gradedqualityDmaintenanceRead-only MCP server that retrieves SEO analytics data from WordPress, Google Search Console, and Google Analytics 4, enabling users to fetch posts, search performance, page metrics, and more via natural language.-
- AlicenseNot gradedqualityDmaintenanceA powerful Model Context Protocol (MCP) server that provides comprehensive Google Analytics 4 (GA4), Google Search Console (GSC), and Google Merchant Center (GMC) integration for Claude Desktop and other MCP clients.Apache 2.0
- AlicenseNot gradedqualityDmaintenanceRead-only MCP server that gives AI clients access to Google Search Console data, enabling natural language queries about traffic, rankings, and SEO opportunities.101 npmMIT