Codex usage dashboard
for your whole team.

How to check your Codex usage

Two commands in the Codex CLI answer it for your own account. To keep your limits in view, add them to the footer with /statusline.

Outside the CLI, the usage dashboard at chatgpt.com/codex/settings/usage shows your current limits and when they reset. Once you reach them, available credits let you keep working; what each plan allows is on Codex pricing.

• /status: the session's model, approval policy, token usage and context left, plus your remaining usage limits.
• /usage: your ChatGPT token activity, or a rate-limit reset if you have one to use. /usage daily, /usage weekly and /usage cumulative open the activity view directly.

Codex usage from scripts, logs and OpenTelemetry

codex exec --json prints one JSON event per line, and each turn ends with a turn.completed event whose usage holds input, cached input, output and reasoning tokens. For past sessions on one laptop, ccusage reads Codex's local logs into daily, weekly, monthly and session reports.

For your own dashboards, Codex exports OpenTelemetry log events: each API request, prompt, tool decision and tool result, with token counts on each completed response. It's off by default. Turn it on with an [otel] table in ~/.codex/config.toml; Codex ignores otel in a project's .codex/config.toml.

codex exec --json "summarize the repo structure" | jq 'select(.type == "turn.completed") | .usage'
bunx ccusage codex daily   # this machine's Codex tokens by day

[otel]
environment = "staging"   # defaults to "dev"
log_user_prompt = false   # redact user prompts unless explicitly enabled
exporter = { otlp-http = {
  endpoint = "https://otel.example.com/v1/logs",
  protocol = "binary",
  headers = { "x-otlp-api-key" = "${OTLP_TOKEN}" }
}}

Codex analytics for a workspace

Workspace admins get two views, and which parts they see depends on role and plan. Workspace analytics (chatgpt.com/admin/usage) covers adoption across ChatGPT. The Admin Console (admin.openai.com) has Codex analytics: Usage for activity and consumption, Insights for what the work was, as use cases and tasks with their messages, credits and active users, and Code review for Codex's reviews.

For automated reports, Enterprise workspaces use the Daily Usage Analytics API, which returns daily ChatGPT, Codex and Work usage for a workspace Admin key scoped to enterprise.analytics.usage.read. Its routes and fields are in OpenAI's Admin API reference.

How Zest tracks Codex usage across a team

OpenAI's views stop at Codex: they don't show the Claude Code or Cursor sessions the same engineer runs, or which sessions ended in a merged pull request. Each engineer installs the Zest plugin for Codex from its marketplace (setup; how plugins work is on Codex plugins). It reads an allowlisted set of session files from ~/.codex, filters them for privacy on the engineer's machine, and syncs them on demand, or in the background once enabled.

• Per engineer: Codex sessions, the model behind each one, and its input, cached and output tokens.
• The Codex skills and MCP tools each session used, listed per engineer next to their Claude Code skills.
• Next to Claude Code, Cursor, GitHub Copilot Chat, Grok Build and Hermes on the same dashboard, so Claude Code vs Codex is measured on your own team's work.
• No code diffs are uploaded.

What Codex usage looks like on our own team

Of the 2,620 pull requests merged in our two repos from July 5 to October 3, 2026, 133 (5%) came from Codex, from 4 engineers. In the 30 days to October 3 the Codex share rose to 9.9%.

866 pull requests merged in our two repos from August 12 to October 4, 2026 carry a session-analytics block, written by the agent that opened them. Codex wrote 31 of them, all in September, which fits the rise above, with a median 576 lines changed against Claude Code's 691.

From Codex sessions to merged pull requests

With the Zest GitHub App connected, each engineer's Codex sessions sit next to the pull requests they merged, with their PR throughput: merged pull requests per week, averaged over four weeks.