How to check your Codex usage
Where OpenAI shows your plan's usage limits and token use, what the percentages measure, and where Codex records them on your computer.
Where can I see my Codex usage?
The Codex usage dashboard shows your current limits and reset times. OpenAI suggests checking it "every week or two to understand your pace and remaining capacity." See OpenAI pricing.
In an active Codex CLI session, /status shows your remaining limits, with the model, approval policy, writable roots and current token usage. /usage opens a menu where you can view account token activity or redeem an available earned reset. /usage daily, /usage weekly and /usage cumulative open token-activity views directly. Without Codex service account authentication, Codex asks you to sign in. /statusline can keep rate limits and token counters in the footer. See Codex CLI slash commands.
In the ChatGPT desktop app, /status shows the chat ID, context usage and rate limits. /usage opens usage and billing settings when available with ChatGPT sign-in. See desktop app slash commands.
Is the OpenAI API usage page the same thing?
No. The API platform's Usage page monitors API-key usage once tracking is enabled, as described in the OpenAI API docs. Codex used with an API key is charged at standard API rates; your ChatGPT plan's limits are on the Codex usage dashboard. See OpenAI pricing.
What do the five-hour and weekly percentages measure?
The Codex app-server docs describe usedPercent as current usage within the quota window, windowDurationMins as its length in minutes and resetsAt as the next reset's Unix timestamp in seconds.
OpenAI gives estimates of local messages per five-hour period for Plus, says these are not fixed message limits, and says Pro plans currently have no five-hour limit. Weekly limits may also apply, and local messages and cloud chats share your plan's usage allowance. See OpenAI pricing.
Model choice, context, reasoning, tool use, retrieval and caching all affect usage, so prompt length alone is not a reliable estimate. Fast mode uses included subscription limits at 2.5 times Standard for the same model. Image generations use included limits 3-5x faster on average than similar turns without image generation, depending on image quality and size. See OpenAI pricing.
For reset times and what a banked or paid reset changes, see how to find your Codex reset time. The Codex reset tracker records reset announcements, and Banked vs automatic Codex resets explains the two kinds.
Where does Codex record usage on my computer?
Codex saves session transcripts under $CODEX_HOME/sessions, by default ~/.codex/sessions. See OpenAI troubleshooting.
In the JSONL session files inspected for this guide, event_msg records with payload.type set to token_count include payload.info.total_token_usage and payload.info.last_token_usage. Their token fields are input_tokens, cached_input_tokens, cache_write_input_tokens, output_tokens, reasoning_output_tokens and total_tokens.
The sibling payload.rate_limits block has windows that, when present, include used_percent, window_minutes and resets_at. The observed window lengths are 300 minutes for five hours and 10080 minutes for a week.
Identify a window by window_minutes, never by whether it is called primary or secondary. In the inspected older CLI builds, the five-hour window is in primary and the weekly window in secondary; newer builds report the weekly window in primary and leave secondary null.
Newer records can also use token_usage_record, with usage, turn_token_usage and thread_token_usage in payload. These are observed session-file keys, not a documented OpenAI file-format contract, and they may change between versions.
How can I inspect my recorded token usage?
The token-counter plugin reads the usage Codex records on your computer and builds a local report of input, output and cached tokens. Ask Codex to "Generate a token report".
Its weekly limit readings are reported by Codex; the plugin does not derive them from token counts. Sharing to this site is optional.
How do I make my Codex usage last longer?
OpenAI suggests precise prompts, less source material, output matched to the need, smaller AGENTS.md files nested within the repository, and turning off MCP servers you don't need, since every MCP server adds context to your messages. If usage is higher than expected, it suggests a smaller model or a tighter task scope. See OpenAI pricing.
For what each plan includes, see Codex pricing: ChatGPT Plus and Pro. To estimate which plan holds your usage, try the plan calculator.
This site is unofficial and not affiliated with OpenAI.