Guides
Guides to monitoring AI coding tools
One page per question, written by the people who built the agent that answers it. Each one says what the tool vendors' own telemetry gives you, quotes their documentation rather than paraphrasing it, and states plainly where our product stops.
Where a vendor's own dashboard is the right answer, these pages say so. For a team that only uses Cursor, it usually is.
How do you monitor Claude Code usage across a team?
Claude Code exports OpenTelemetry metrics, events and traces. What you get from it, what it costs to run for a whole team, and where per-user attribution breaks.
Read the guideHow much is your team spending on AI coding tools?
Three different numbers get called AI coding spend: seats, metered API billing, and tool-reported cost. Where each comes from, and how to get cost per developer out of them.
Read the guideWhat the Cursor team dashboard and Admin API do not show you
Cursor's usage analytics is good at Cursor. The documented limits: 10 users and 90 days of filtering, no Cursor CLI or Background Agents, and on-device code scoring.
Read the guideHow do you track Codex CLI usage across a team?
Three routes: OpenAI's enterprise Analytics API, Codex CLI's own OpenTelemetry config, or reading the telemetry on each machine. What each one covers, and what it misses.
Read the guideWhat happens when an API key goes into an AI coding prompt?
The prompt has already left the machine, so this is detection and not prevention. What on-device scanning can tell you, what it misses, and what to do in the first hour.
Read the guideHow do you measure AI coding tool adoption?
Seats assigned is not adoption. The two numbers that are, the signals each tool emits to support them, and five ways teams measure this wrong.
Read the guideDo you need an OpenTelemetry stack to monitor Claude Code?
No, you need something that accepts OTLP. The three shapes that takes, the cardinality trap in the default attributes, and when building it yourself is the right call.
Read the guideCan you monitor AI coding tool usage under GDPR?
Lawful basis, data minimisation, and the works council conversation. What the design decisions are, what an aggregate-only mode does at the storage layer, and what is missing.
Read the guideCan you keep AI coding tool usage data on your own servers?
Two routes: run the product yourself, or collect the tools' OpenTelemetry into your own store. The checks that separate real self-hosting from an on-prem label.
Read the guideWhich AI coding tools are your developers using?
An endpoint agent gives you an inventory of the tools it supports, not a discovery of the ones you have never heard of. What each approach answers, and where to look for the rest.
Read the guide