GitHub CLI begins collecting pseudoanonymous telemetry

GitHub CLI (gh) has begun collecting pseudonymous telemetry.

The purpose of the collection is to understand feature usage as agentic usage increases, and to prioritize based on actual usage patterns. The intent is to check data such as adoption rates of new subcommands and how frequently certain flag combinations are used, in order to improve feature discoverability and UX.

The items collected include the following.

• Command execution info: command, flags, version
• Environment info: os, architecture, ci, github_actions, is_tty
• Identifiers: device_id (a locally stored UUID), invocation_id (a per-invocation UUID)
• Agent/skill related info: agent, agent_hosts, skill_names, skill_repo, etc.

Telemetry is sent to GitHub's internal analytics infrastructure. However, users can check exactly what payload is generated before it is sent.

There are two ways to check.

• Environment variable: GH_TELEMETRY=log
• CLI config: gh config set telemetry log

In log mode, the JSON payload is printed to stderr instead of actually being sent.

Disabling is also possible.

• GH_TELEMETRY=false
• DO_NOT_TRACK=true
• gh config set telemetry disabled

The environment variable takes precedence over config. Also, GitHub CLI extensions and Copilot CLI may have separate data collection policies, so this setting alone does not turn off all data collection.