llms.txt
Analytics — Free AI Data Metrics
You shipped an agent, an MCP server, a CLI, a VS Code extension. You have no idea whether anyone runs it twice. Fetch one URL with your account ID and you have usage analytics — no SDK to install, no API key to exchange, no JSON to assemble, nothing to add to your dependency tree.
tool_called
Four you will use, a few you might, and anything else you add is kept as an attribute on the event. There is no schema to extend.
| Parameter | Meaning |
|---|---|
| uid | Your account ID. |
| c · cc · s_p · n | Your user's city, county, state/province and nation. Required: a ping's own address is where your software runs, not where the person is. |
| event | What happened — page_view, tool_called, run_completed. Anything you like. |
| project | Which of your tools this is. Keeps them separate under one account. |
| type | Channel. ai for an AI tool; that is the default here. |
| sid | Session ID. Pings that share one are grouped into a run or conversation. |
| path | A screen or step. With event=page_view it becomes a real pageview. |
| anything | Kept as an attribute: &tool=search&latency_ms=420&outcome=success |
Event names are yours to choose. These are the ones that answer the questions you actually have about a tool you just shipped.
Is anyone adopting it, and are they staying current? Pass version=.
How many runs finish. Pass outcome=success|error|cancelled.
Which capabilities actually get used. Pass tool= and latency_ms=.
Which screens or steps people reach, with path=.
What breaks in the wild. Pass kind= — never the message text.
Whether it helped. Pass rating=up|down.
Hand one line to your coding agent. It reads the contract, opens an account, wires up the events, and updates your own llms.txt so the next agent does the same.
llms.txt with the instructed changes.
Read llms.txt
— the whole contract. No signup form, no key exchange.
/llms.txt
the contract, read first
/api/v1/accounts
account created
/api/ping
run_started · tool_called · run_completed
Every parameter checked against the do-not-send list before it went in a URL.
Copy this and give it to your coding agent:
Read https://seriouslysimpleanalytics.com/llms.txt and instrument this project with SeriouslySimpleAnalytics. Create an account and set project to this tool's name. Report at least: first run, run started/completed with an outcome, each tool call, and errors by kind. Reuse one sid for the whole run. Never send credentials, prompts or completions. Update our llms.txt with the instructed changes.