GitHub has started collecting pseudonymous client-side telemetry from its command-line interface by default, with an opt-out. The change affects GitHub CLI starting with v2.91.0 and was introduced through documentation updates, release notes, and code changes in the CLI repo.
The company says the data will help it understand how people use GitHub CLI, spot friction, and decide which features deserve more work. That pitch is familiar across the software industry: usage data can make products better, but default-on telemetry is also how trust gets spent.
What GitHub CLI telemetry collects
GitHub says the telemetry begins with the v2.91.0 release and is enabled by default with an opt-out. The public documentation does not give a full inventory of the fields collected, which is exactly the sort of omission that makes privacy-minded users reach for the nearest configuration file.
In the sample payload GitHub provides, the data can include:
- Agent field
- Architecture
- Device ID
- Operating system
- Flags
- Command name
- Invocation ID
- Other metadata
GitHub also says the exact payload may differ, so the sample is more of a hint than a promise.
How to switch telemetry off
The company does at least provide an off switch. Users can disable telemetry by setting GH_TELEMETRY=false or DO_NOT_TRACK=true in the environment, or by running gh config set telemetry disabled in the CLI.
That matters because the easiest privacy setting to defend is usually the one you never have to hunt for. GitHub’s CLI is open source, which means anyone can inspect the implementation, but open code is not the same thing as clear communication – especially when the default setting is to collect data first and explain later.
GitHub’s focus on agentic workflows
The tell here is not just telemetry, but the reason GitHub gives for it. The company explicitly ties the change to ”agentic adoption” of GitHub CLI, which suggests it wants visibility into both human developers and AI agents using the tool. That is a useful clue about where GitHub thinks the platform is headed: not just terminal power users, but software acting on their behalf.
Competitors are making similar moves across developer tools, where usage analytics and AI features increasingly arrive as a pair. The pattern is straightforward: vendors want data to tune products, and AI products need even more data to prove they are being used. The result is a neat little loop – useful for product teams, less charming for everyone else.
The immediate question is whether GitHub keeps the current level of opacity or publishes a clearer list of telemetry fields. If the company wants developers to trust its CLI while it leans harder into AI-driven workflows, it may need to explain exactly what ”pseudonymous” means before the next default setting quietly changes too.