Week over week
Compare this week's traffic with last week's for every site.
One table across properties.
Reading GA4 from Gemini CLI usually starts in the Google Cloud console. With LopeBase you connect with Google once, run one command, and Gemini CLI can read traffic, channels, landing pages, and conversions for every property you run.
Read only. Nothing to paste into settings.json.
One question, one tool call, and an answer built on your own GA4 numbers.
How did traffic change this week compared with last week, for all my sites?
get_digest
| property | this week | last week | change |
|---|---|---|---|
| acme.com | 18,420 | 16,990 | +8% |
| acme.shop | 6,210 | 7,480 | −17% |
| acme.dev | 3,105 | 3,020 | +3% |
Two sites are steady or up. acme.shop fell 17%. Want me to read its traffic sources and see which channel dropped?
No project, no enabled APIs, no credentials file. Connect with Google in LopeBase covers it.
Compare all your GA4 properties in one answer, with the digest or per property.
Type /mcp in Gemini CLI to see LopeBase and its tools listed.
Three steps. The first is the same for every app; the rest is specific to Gemini CLI.
Connect GA4 to LopeBase. Start free, add your site, press Connect with Google, and pick the GA4 property. You need no Google Cloud project and no API key. Google is only asked for read access.
Run npx lopebase in your project folder. It signs this computer in and registers LopeBase with Gemini CLI, so you never paste a token into settings.json.
Start Gemini CLI and check. Type /mcp to see LopeBase and its tools listed, then ask your question.
$ npx lopebase
Works on macOS, Linux, and Windows with Node 18 or later. The same command can also set up Claude Code, Codex, opencode, or Cursor Agent if you use more than one.
Quick checks that fit in a terminal between other work.
Compare this week's traffic with last week's for every site.
One table across properties.
acme.shop traffic fell. Which channel lost the most visitors?
Narrow a drop to its source.
What were the ten most viewed pages in the last 7 days?
A quick look before you write or ship.
List the goal events and how many people triggered each in the last 7 days.
Make sure tracking still works after a deploy.
Seven datasets per GA4 property. Breakdowns cover the last 28 days and the trend up to 90. Pass any event as a goal to count conversions.
More on each dataset: Google Analytics MCP. Connecting in the console: the GA4 connector.
LopeBase is not the only route. Here is how it compares with the usual one.
| Topic | Google's Analytics MCP server | LopeBase |
|---|---|---|
| Setup | Python and pipx, a Google Cloud project, two Analytics APIs, local credentials, and an entry in settings.json. | npx lopebase and Connect with Google. |
| Credentials | A credentials path and project ID in your settings. | Kept by LopeBase, encrypted. |
| Flexibility | Any GA4 report you describe, including custom dimensions. | Seven shaped datasets, with goals built in. |
| Other data | GA4 only. | Search Console, Bing, PostHog, Stripe, and uptime too. |
| Best when | You already work in Google Cloud and need custom reports. | You want GA4 in Gemini CLI in a few minutes. |
Google's server is described from its own README, checked October 2026. It is labeled experimental there, and its README covers Gemini CLI setup.
Something else? Contact us.
Yes. They are separate MCP servers. Use Google's for custom reports and LopeBase for the everyday questions and your other sources.
With Gemini CLI's own MCP settings, as npx lopebase mcp. No token goes in settings.json.
This page is about Gemini CLI. LopeBase works in any client that can add a remote MCP server with sign in.
No. It gets counts and rates, never a visitor, an IP address, or a user ID.
No. You pick the property when you connect with Google, and Gemini CLI asks by name.
Free for one product. No card. Agent access is on every plan.