First request

Verifying captures are working

After completing a real API call (not just --help):

halton-meter report

A row appears for every intercepted LLM request. Columns include: timestamp, project, model, tokens in/out, cost.

If report is empty but the daemon is healthy, work through this checklist:

1. Check which mode you installed with: halton-meter status — look for the install-mode row
2. In env-only mode, only halton-meter run <cmd> calls are metered — a bare claude in your terminal is not captured
3. In --apps mode, terminal-launched calls need a new terminal (so the rc export takes effect); IDE-launched calls need the IDE to be restarted after init
4. See No captures in halton-meter report in the troubleshooting guide for all six known causes

Daily usage

env-only mode

Prefix any command with halton-meter run:

$ halton-meter run claude                    # meter Claude Code
$ halton-meter run python my_script.py       # meter an SDK script
$ halton-meter run -- npx my-llm-tool       # use -- to pass flags
$ halton-meter run --shell                   # interactive metered subshell

apps or full mode

After a new terminal or IDE restart, your env vars are already set. Run your tools normally:

$ claude                        # captured automatically
$ python my_script.py           # captured if it uses requests / httpx / urllib

You can still use halton-meter run explicitly in any mode — it is always safe to combine.

Checking daemon health

$ halton-meter status           # quick health summary
$ halton-meter doctor           # full diagnostic with copy-paste fixes

Starting and stopping

$ halton-meter start                              # (re)start the daemon via launchd
$ halton-meter stop                               # stop the daemon
$ halton-meter stop && halton-meter start   # restart after config change

Switching install modes

You can switch modes without uninstalling first. Just re-run init with the mode you want:

$ halton-meter init             # switch to env-only
$ halton-meter init --apps      # switch to apps
$ halton-meter init --full      # switch to full

The reconciler cleans up all prior-mode state — stale launchctl env vars, shell rc blocks, system proxy settings — so the new mode is the sole source of truth. You do not need to run uninstall first.

Once captures are landing on this machine, the next question is usually running meters on more than one machine — workstations, CI runners, a teammate’s laptop — and rolling them up into one report.

Running on more than one machine? →