Kibble Blog

Codex usage dashboard: check limits, tokens and team usage

By Kibble · 5 September 2026 · 5 min read

To check Codex usage, open OpenAI's usage dashboard for your account limits, or type /status inside Codex CLI. Use /usage for account token activity. To track supported local usage across a team's machines and agents, connect a collector such as Kibble. Its Codex costs are API-rate estimates; they do not tell you what OpenAI billed.

How do you check Codex usage in the CLI?

Start an interactive Codex CLI session, then enter /status in its chat input. It shows session configuration and token usage; OpenAI also documents it as the way to check remaining usage limits during a session.

/status
/usage daily
/usage weekly
/usage cumulative

These are commands inside Codex, not shell commands. /usage opens the account usage menu; the three forms above open a token-activity view directly. If Codex asks you to sign in, that view requires supported account authentication. Command availability can differ on older releases. See the official OpenAI slash-command reference, checked 22 September 2026.

Which Codex usage dashboard should you use?

Choose the view by the question: remaining allowance, account history and team activity are different measurements.

OpenAI's workspace analytics guide separates interactive Codex reporting, aggregate Analytics API data and compliance records. Check what your workspace already includes. Kibble adds a combined view across supported agents and machines; it does not replace those controls or audit records.

Does a Codex usage tracker show your monthly bill?

Kibble's local token-cost estimate does not establish your subscription bill, credit balance or remaining allowance. Keep those figures separate when reviewing a month.

OpenAI's pricing and usage guide directs you to its usage dashboard for current limits and reset times. API-key usage is charged under API pricing. For actual charges, use the billing records for the account that paid.

Kibble estimates local token cost using model rates. That helps compare recorded activity across days, models and repositories, but a subscription can include that work. An estimate above a subscription price does not prove an overage or a saving. For a monthly review, select the same calendar dates in each report and keep tokens, estimates and actual charges in separate columns.

How do you connect a machine running Codex?

Use Node.js 20 or later on a machine with Codex session logs. First check that the collector can find them:

npm install -g @usekibble/cli
kibble doctor
kibble login
kibble push
kibble usage --range week
kibble schedule status

kibble doctor checks local collection before login. kibble login opens the browser to sign in with Google and links that machine to your own member identity. It sends available usage immediately when automatic collection is on, then schedules hourly and startup pushes. Initial sync includes today and the preceding 29 UTC days where logs exist. If automatic collection is off, run kibble push manually. Repeat the setup on each machine whose local work you want included. The setup guide covers policy settings and status commands.

The free personal plan covers one person on two machines with 30 days of history. For a team, an owner upgrades to Pro, invites engineers at their verified email addresses and assigns teams. Check the team price and trial terms before upgrading. Every engineer links their own machines. Sharing one person's login would make the per-engineer breakdown misleading.

What does the Codex collector see?

The implementation is public in Kibble's source adapters. The collector reads supported JSONL records under $CODEX_HOME/sessions and $CODEX_HOME/archived_sessions; CODEX_HOME defaults to ~/.codex. It reads token counters and derives repository names locally. Repeated cumulative token snapshots are deduplicated, rather than added together. Cloud work or other activity without supported logs on a linked machine is outside this local view.

The owner can see member totals. A manager can see assigned teams, and a member sees themselves. Device details remain owner-only. Names, counts and allowed billing-plan metadata leave the machine; prompts, code, tool arguments and paths do not. The privacy policy gives the full field boundary.

Why is Codex usage missing from the tracker?

Check the machine, log location and reporting dates first. Run kibble doctor where Codex actually ran, then kibble push and kibble schedule status. A successful login alone does not prove that supported usage was found.

If you use a custom CODEX_HOME, check that collection runs with that location. Compare a known day of local Codex activity with the same dates in Kibble. Missing logs, unsupported formats and work done only in the cloud must stay unknown. Never interpret an empty row as proof that an engineer did no work.

Why is a Codex skill's invocation count zero?

Kibble's Codex adapter counts explicit skill selections recorded in supported user-message entries and recognized command submissions in local CLI history. It also inventories supported skill directories. Automatic skill reads and skill cost attribution remain unobserved; commands missing from local history are not counted. A zero invocation or attributed-cost value therefore cannot establish that the skill was unused or free.

Ask the skill's owner about its purpose and check the agent's configuration locally. The skill cleanup method explains why the evidence differs between agents.

How do you read the team's weekly figures?

1. Check which members reported and whether the scheduled pushes are healthy. Missing data is not zero usage.
2. Compare the same date range by agent and model. A model change can change estimated cost without the same change in tokens.
3. Use repository estimates to identify a change worth discussing, not to allocate the vendor invoice to the cent.
4. Check subscription or API billing before interpreting equivalent API cost as money spent.

The personal plan retains 30 days; Pro retains all history received. Kibble cannot recover deleted local logs that were never collected. For the underlying log format and cumulative token-counter caveat, read where Codex keeps its usage logs.