2am, prod is on fire. No browser, no VPN to the dashboard.
Agent queries error rate via HogQL, ramps the flag if green. No human in the loop.
Person timeline, flag variants, recent events. All from a shell pipeline.
Check flag state before merging. Fail the pipeline if it's not ready.
Find which model, prompt, or user is burning your LLM budget from the terminal.
LLM writes the HogQL, bosshogg runs it, agent interprets results. No API key juggling.
$ brew install aaronkwhite/tap/bosshogg
$ cargo install bosshogg
$ bosshogg login # browser device-flow
$ bosshogg doctor # verify everything
$ bosshogg configure # wizard: host + API key