cogny clianalyticsfirst 15 calls free
google analytics 4 queried from the terminal.
Run GA4 reports and realtime reports, list dimensions, metrics, conversion events and audiences, and build audiences — from Bash over the GA4 MCP. The Data API without the boilerplate.
- 1. sign up, free for the first calls (uses your git email)
npx @cogny/cli init --agent claude-code - 2. see every Google Analytics 4 tool
cogny tools list - 3. call any tool from bash, no MCP wiring
cogny tools call mcp__Google_Analytics_4__tool_run_report --input '{…}'
The GA4 Explore UI is slow and the Data API needs a service account and client library. The CLI gives you the same reports as one JSON call you can schedule or pipe. Connect Google Analytics 4 once in Cogny Solo and let your coding agent, or a scheduled loop, do the querying for you.
Start SoloSign yourself up with one command. The CLI issues a key, wires nothing, and exposes every Google Analytics 4 tool over JSON-RPC. The first 15 calls are free.
npx @cogny/cli init --agent claude-code
real Google Analytics 4 tools.
Tool names as the Google Analytics 4 MCP serves them. Send --input JSON, get JSON back. The payloads are illustrative.
- mcp__Google_Analytics_4__tool_run_report
A GA4 Data API report: dimensions × metrics over a date range.
cogny tools call mcp__Google_Analytics_4__tool_run_report \ --input '{"property_id":"353328203","dimensions":["sessionDefaultChannelGroup"],"metrics":["sessions","conversions"],"date_ranges":[{"start_date":"7daysAgo","end_date":"today"}]}'
example response
{ "rows": [ { "sessionDefaultChannelGroup": "Organic Search", "sessions": 18422, "conversions": 412 }, { "sessionDefaultChannelGroup": "Paid Search", "sessions": 9211, "conversions": 388 }, { "sessionDefaultChannelGroup": "Direct", "sessions": 7044, "conversions": 121 } ] } - mcp__Google_Analytics_4__tool_run_realtime_report
Users on the site right now, by dimension.
cogny tools call mcp__Google_Analytics_4__tool_run_realtime_report \ --input '{"property_id":"353328203","dimensions":["unifiedScreenName"],"metrics":["activeUsers"]}'
example response
{ "rows": [ { "unifiedScreenName": "/pricing", "activeUsers": 23 }, { "unifiedScreenName": "/", "activeUsers": 17 } ] } - mcp__Google_Analytics_4__tool_list_conversion_events
Which events are marked as conversions.
cogny tools call mcp__Google_Analytics_4__tool_list_conversion_events \ --input '{"property_id":"353328203"}'
example response
{ "conversion_events": [ { "event_name": "purchase", "counting_method": "ONCE_PER_EVENT" }, { "event_name": "sign_up", "counting_method": "ONCE_PER_SESSION" } ] }
the rest of the surface.
A sample of the other Google Analytics 4 tools. Run cogny tools list --json for every tool and its input schema.
- mcp__Google_Analytics_4__tool_list_account_summaries
- mcp__Google_Analytics_4__tool_get_metadata
- mcp__Google_Analytics_4__tool_list_custom_dimensions
- mcp__Google_Analytics_4__tool_list_custom_metrics
- mcp__Google_Analytics_4__tool_list_audiences
- mcp__Google_Analytics_4__tool_create_audience
- mcp__Google_Analytics_4__tool_list_data_streams
- mcp__Google_Analytics_4__tool_list_google_ads_links
calls from a script. verdicts from cogny.
Every write is logged with who authorised it. Launches your agent makes (a new campaign, a new ad) become measured tickets automatically; for other changes, your agent offers to register them. Cogny measures each in your own data, three days later for ads and seven for content and SEO, and closes it as worked, partial, didn't work or inconclusive.
- worked
Moved the number. Every measured win has a share card.
- partial
Some of it did.
- didn't work
On the record anyway, so it isn't proposed again.
- inconclusive
Not enough data to call.
questions.
- Is this the GA4 Data API?
- It wraps it. `tool_run_report` maps to a runReport call; `tool_run_realtime_report` to a realtime report. You get the rows as JSON without a service account or the google-analytics-data client.
- Can I schedule these from cron?
- Yes. The CLI is a normal binary — wire `cogny tools call tool_run_report --input ...` into cron, a Makefile, or your agent and capture stdout.
- Does it cost a credit per call?
- No, on every paid plan (Solo, Cogny AI, Cogny Pro and Cloud) standard MCP calls are unmetered under fair use. The Free plan includes 15 MCP calls a month.
Building an agent? The full tool map for every server is one fetch: /cli/llms.txt