A google search console mcp app lets your AI assistant read your own Search Console data and answer questions about it in plain English - no dashboard, no date picker, no spreadsheet export. Install a hosted app, connect it to Claude (or ChatGPT or Cursor), and instead of building a report you type one question and read the reply. This article walks through two apps that do exactly that - Google Search Console for how you show up in Google, and Web Analytics for what visitors do once they arrive - with copy-pastable prompts, setup steps and the honest limits of each.
- What it is: a hosted MCP app that lets Claude query your Search Console and Umami analytics using your own credentials.
- Two apps: Google Search Console (read + write) for search performance and indexing; Web Analytics (read-only) for visitors, referrers and devices.
- Setup: paste a connection URL into Claude's custom connectors, sign in with OAuth - no API key, no Google Cloud project.
- Use it for: click-loss diagnosis, rising queries, CTR drops, indexing checks, traffic digests, and combined search-plus-analytics summaries in one turn.
- Know the limits: Search Console data lags 2-3 days and goes back 16 months; Umami is cookie-free and is not GA4.
Why ask Google Search Console in plain English (and what MCP has to do with it)
B77 is an app store for your AI assistant, and a Google Search Console MCP app is the piece that lets your assistant read your own Search Console data. Think of it the way you think about a phone: it ships with built-in apps, and you install more to extend what it can do. Claude, ChatGPT, and Cursor ship with built-in abilities, and B77 apps extend them with your tools and your data.
The connective tissue here is the Model Context Protocol (MCP), the open standard that lets an AI assistant call external tools. Until you install a Google Search Console MCP app, Claude simply cannot see your Search Console account or your website traffic. Once you install it, the work changes shape. Instead of building a report, you ask a question.
Compare the two routines. The old one: open the dashboard, set the date picker, filter by page or query, export to a spreadsheet, then pivot until the answer appears. The new one: type one question in the chat and read the reply. This article uses two apps to do exactly that: Google Search Console (B77) for how you show up in Google, and Web Analytics (B77) for what visitors do once they arrive. If you own a website and want answers rather than analytics dashboards, that is the point.
What the two apps can see and do
These two hosted MCP servers each expose a focused set of tools. The facts below are as of 2026-09-01, from the app pages on b77.ai.
Google Search Console app
This Google Search Console MCP server is hosted at B77. It is free and works in Claude, ChatGPT, and Cursor, with both read and write access. Its tools cover performance reports - clicks, impressions, CTR, and average position across queries, pages, countries, devices, and dates - plus comparing any two periods and branded-vs-non-branded query splits and query opportunities. On the indexing side, you can list and submit sitemaps and run URL inspection to check a URL's Google index status. It reads Search Console performance and indexing data for the properties you connect. It never sees your messages, files, or other apps, and it does not keep the data a request returns: each request queries Google live using your own credentials, while sign-in tokens sit in a per-user vault and every tool call lands in the audit log.
Web Analytics app
The Web Analytics MCP server is free and read-only. It is built on Umami, an open-source, cookie-free analytics tool under the MIT license. Its tools answer visitors, pageviews, top pages, referrers, countries, browsers, and devices; give a full-site summary; track one page over time; and filter by country and by how many days back. It reads aggregate traffic stats only and sees no personal visitor data.
| Search Console app | Web Analytics app | |
|---|---|---|
| Data source | Google Search Console (your credentials) | Umami (open-source, cookie-free) |
| Access | Read + write | Read-only |
| What it answers | Search performance, indexing, sitemaps, URL inspection | Visitors, pageviews, referrers, countries, devices |
| What it never sees | Your messages, files or other apps; no response data kept | Personal visitor data |
Connect both apps to Claude in five minutes
These steps were verified on 2026-09-01 against the B77 app page and the Claude Help Center article on custom connectors. Claude connectors are the links Claude keeps to remote MCP servers such as these two apps.
- Open the Google Search Console app page on B77 and click Install. Copy the connection URL:
https://google-search-console.b77.ai/mcp. - In Claude, go to the main menu → Customize → Connectors → Add → Add custom connector.
- Give it a name, paste the URL into Remote MCP server URL, and sign in with your B77 account.
- Repeat with the Web Analytics app:
https://web-analytics.b77.ai/mcp.
Like other Claude MCP connectors, it then appears in claude.ai, Claude Desktop, and Claude Code. Expected result: Claude lists the tools and asks for approval the first time it calls one. Per the Claude Help Center, click Allow always only for a server and tool you trust to run unsupervised.
Per the Claude Help Center (checked 2026-09-05), custom connectors using remote MCP are available on Free, Pro, Max, Team, and Enterprise plans; free users are limited to one custom connector. On Team and Enterprise, an owner adds it under Organization settings → Connectors, and members then see it under Customize → Connectors.
To connect in ChatGPT: Settings → Security and login, enable Developer Mode, then Plugins → + and paste the URL. In Cursor, add the URL in the MCP settings or your mcp.json.
OAuth sign-in (the standard that grants an app scoped access without sharing your password) means each app runs with the credentials you authorize for its own data source - your Google account for Search Console, your analytics account for Web Analytics - and B77 never sees your passwords. There is no API key to create and no Google Cloud project to set up on your side.
Six Search Console questions a PM asks every week (1-6)
With both apps connected, here are the questions that used to mean opening Search Console, picking a date range, adding a filter, and exporting to a spreadsheet. Now you type them as prompts and read a table. Below each prompt is what a good answer looks like and what to do with it.
1. Which pages lost clicks compared with the previous period?
Using Google Search Console for https://example.com/, compare the last 28 days
with the prior 28 days by page. Show clicks, impressions and CTR for both
periods and the click change. Return a table sorted by largest click loss first.Look at the top rows: those are the pages bleeding the most clicks. Check whether impressions also fell (a demand or ranking issue) or held steady while clicks dropped (a CTR issue). Pick the worst one or two and investigate before touching anything.
2. Which queries are rising?
Using Search Console for https://example.com/, group by query and compare
impressions this week versus last week. Show queries with growing impressions,
with clicks, impressions, CTR and position. Return a table sorted by impression gain.Rising impressions mean Google is showing you for more searches - often before clicks catch up. Scan for queries you don't yet have a dedicated page for; those are content opportunities. Note any where position is weak but impressions are climbing.
3. Where did CTR drop while position held?
Using Search Console for https://example.com/, compare the last 28 days with the
prior 28 days by page. Show pages where average position stayed roughly the same
but CTR fell. Return a table with both periods' clicks, impressions, CTR and position.Stable average position with falling CTR can point at a snippet issue, but check the query mix and the SERP first - new features or competitors can cut CTR while position looks unchanged. If the snippet is the culprit, rewriting the title and description is the fastest win on the list.
4. Is the page I published yesterday indexed?
Run URL inspection on https://example.com/new-page/ for the property
https://example.com/. Tell me whether it is indexed, and if not, why.
If it is not indexed, explain the reported status and what to fix first.The inspection result tells you if the URL is on Google, was crawled, or is excluded - and names the reason when it is not indexed, which tells you what to fix first. New pages can take Google a few days to pick up either way.
5. Are pages stuck outside the index or dropped from the sitemap?
For https://example.com/, list the submitted sitemaps and their last-submitted
status. Flag any errors. Then help me submit an updated sitemap URL.Check the last-submitted date and error count per sitemap - a stale or erroring sitemap can slow discovery of new URLs, even though Google can still find them through internal links. If your sitemap URL changed, submit the current one. Fix errors at the source, then resubmit.
6. How does a country or device segment perform?
Using Search Console for https://example.com/page/, break down the last 28 days
by device (mobile vs desktop). Show clicks, impressions, CTR and position for each.
Return a table.Compare mobile against desktop on the same page: a big gap by device is a signal to investigate - check the device-specific query mix and SERP first, then mobile usability and speed if mobile still looks weaker. Swap "device" for a single country to spot a market where you rank but don't convert clicks.
Tip: be specific about the property, the date range, and the dimension in every prompt, because the assistant passes exactly those values to the Search Console API - vague queries get vague tables.
Four traffic questions for the Web Analytics app (7-10)
The remaining four questions point at the Web Analytics app, which reads your Umami traffic data. You can name a site by its full domain or by a short alias you have set up - and if you are not sure which sites the app can reach, just ask it to list them first.
List the websites this analytics app can see.7. Give me the weekly traffic digest
For example.com, give me a weekly digest: total visitors and pageviews for the last 7 days versus the 7 days before, the top 10 pages, and the top referrers.Read the two 7-day windows side by side - a rising visitor count with flat pageviews means fewer pageviews per visitor; check engagement separately before calling sessions shorter. The top pages and referrers tell you what carried the week, which is the same summary you would otherwise assemble by hand from a dashboard.
8. Where does the traffic come from?
For example.com, show the top referrers for the last 30 days, and for each referrer the pages people land on.This pairs each source with its landing pages, so you can see whether a referrer sends people to your homepage or deep into a specific article. Direct and "none" entries mean the referrer was not recorded, so treat them as an unknown bucket rather than a real channel.
9. How is one page doing over time?
For the path /pricing on example.com, show visitors by week for the last 90 days.A weekly view smooths out the day-to-day noise you get in raw traffic and makes a trend easier to spot. Watch for step changes that line up with a launch, a campaign, or a change to the page itself - the analytics tell you when something shifted, not why, so pair this with what you know happened.
10. Which countries and devices do visitors use?
For example.com, show the country and device split for visitors over the last 30 days.,
The country breakdown helps you sanity-check whether your audience matches where you actually sell or publish. The device split is a quick prompt to test your key pages on mobile if a large share of traffic is arriving that way - a detail that is easy to miss in a desktop dashboard.
Put the two apps together
Search Console tells you how people found a page - the queries, impressions and clicks that got them there. Web Analytics tells you what happened after they landed - visitors, referrers and which channels sent them. On their own, each answers half a question. In one Claude conversation, you can ask both at once and let the assistant compare the two views.
The first pattern separates a search problem from a site-wide one:
For the three pages that lost the most clicks in Google Search
Console over the last 28 days, pull visitors and top referrers
from Web Analytics for the same period. Tell me whether the
loss is search-only or across all channels.The second is a Monday digest you can paste into a team channel:
Give me one combined summary for last week:
- search clicks and impressions, week over week
- top rising queries
- visitors and top referrers from analytics
- any of the URLs inspected this week that returned an indexing problem
Format as five bullet points for a team channel.This works because both apps are installed in the same assistant, so a single question can call tools from Search Console and from Web Analytics in the same turn. Claude reads both sets of results, lines up the same date range, and writes one answer instead of leaving you to stitch two dashboards together by hand.
Scope honesty: Umami is not GA4
Before you ask your first question, be clear about what you are actually querying. The Web Analytics app reads an Umami instance - an open-source (MIT license), cookie-free analytics tool that collects no personal visitor data. Through the B77 app, it reports pageviews, visitors, referrers, locations, and devices. That is the whole picture.
What it does not give you are GA4 concepts: there are no GA4 audiences, no engagement rate, no ecommerce revenue reports, and no attribution models. The B77 app exposes traffic counts only, so treat it as a way to ask about visits and where they came from, not conversions or spend.
If you need GA4 data inside an assistant, Google publishes its own Google Analytics MCP server (github.com/googleanalytics/google-analytics-mcp, Apache 2.0 license). This GA4 MCP server is read-only, runs locally on your machine, and uses the Google Analytics Admin API and Data API; its developer page was last updated 2025-09-16.
So pick based on the data you have. If you run Umami, the Web Analytics app fits. If your numbers live in GA4, run Google's server. There is no need to compare the two on quality - they simply read different sources of analytics.
Limits and freshness: what the numbers can and cannot tell you
Before you trust an answer, know what the data can honestly tell you. The facts below come from Google's documentation, checked 2026-09-01.
Freshness. Search Console data normally becomes available 2-3 days after it is collected. So if you ask about yesterday, expect preliminary or empty results. The API exposes a dataState parameter that distinguishes final data from fresh data that can still change - worth mentioning when you ask the assistant for very recent numbers.
How far back. Historical performance data goes back 16 months. Year-over-year comparisons work; a two-year lookback does not, because the older data is gone.
Why totals don't add up. To protect privacy, Google does not show some rare or sensitive search queries. Query-level totals therefore never sum to the page total. This is expected, not a bug - flag it if the assistant tries to reconcile the two.
Depth. The Search Console dashboard shows at most 1,000 rows per table. The API accepts a rowLimit up to 25,000 per request, so the assistant sees far more of the long tail - though Google returns top rows and does not promise every last query.
Quotas. The API is metered. Search Analytics allows 1,200 queries per minute per site and per user. URL Inspection allows 2,000 queries per day and 600 per minute per site - so asking the assistant to inspect every URL on a large site in one pass will hit the limit.
Web Analytics. Umami is cookie-free, so its visitor counts are not directly comparable with cookie-based tools. Read the trend, not the absolute number.
Read-only by default: sharing the apps with your team
Handing these apps to a team is safe because of what each one can actually do. The Web Analytics app is read-only: the assistant can look things up and report back, but it never changes anything. There is no write path, so there is nothing to break.
The Search Console app can write - submitting a sitemap, for example. Each write call needs your approval in Claude before it runs, unless you have chosen "Allow always" for that tool. Write actions are also bound by Google's own quotas and by your permission level on the property, so the app can never exceed what your account is allowed to do.
Sharing works from the app's Manage page on B77. You invite teammates by email; they add the same connection to their own assistant and use it under your access, without setting up their own credentials. You can revoke a share at any time, and every tool call is written to the usage log. Access runs through per-user OAuth sign-in scoped to your account; B77 keeps your sign-in tokens in a per-user vault but not the data requests return, and hosting is in the EU (Frankfurt and Helsinki).
Practical advice: give the read-only Web Analytics app to the whole team, keep the Search Console app with the people who own the property, and never paste tokens or passwords into a prompt.
Frequently asked questions
Do I need a Google Cloud project or an API key?
No. You connect a Google account that already has access to the properties you want to query. The hosted app handles the Search Console API calls and the OAuth sign-in for you. There is no project to create and no key to paste anywhere.
Does the Google Search Console AI setup work in ChatGPT and Cursor too?
The same remote MCP endpoint works in compatible MCP clients, but the setup flow and plan requirements differ between ChatGPT, Claude, and Cursor. In ChatGPT, you add it through Developer Mode plugins; in Cursor, you point to it from MCP settings. The questions you ask stay the same - only the place you paste the URL changes.
Is the Web Analytics app Google Analytics?
No. The Web Analytics app reads from Umami, an open-source analytics tool, not GA4. If you need Google Analytics data, use Google's own local MCP server for GA4 instead.
Can the assistant change or delete anything on my site?
The Search Console app's write actions cover sitemap submission; each one sits behind a tool approval and Google's own quotas, and nothing it does can edit your website itself. The Web Analytics app is read-only. This keeps the setup secure by default.
How fresh and how far back is the data?
Search Console data typically lags two to three days and keeps up to 16 months of history. Web Analytics numbers are live, read straight from your Umami instance. Keep both windows in mind when you compare a recent spike against a longer trend.
Next step: install the two apps and ask your first question
B77 works like an app store for MCP servers: you find a hosted tool in the B77 catalog and connect it to Claude in a couple of clicks, with auth handled for you. If you only do one thing this week, ask the click-loss question - compare the last 28 days against the previous 28 and see which queries and pages dropped. That single prompt usually surfaces something worth acting on.
To try it, install the Google Search Console and Web Analytics apps from the B77 marketplace and connect them to Claude. Never signing up is fine too: the same comparison runs by hand in the Search Console dashboard - it just takes longer. Browse hosted MCP servers and connect one to your assistant in a couple of clicks: the B77 marketplace.
