The Codex limit for Plus users, and when a team should pay for more
Plus suits a few focused sessions a week. Here is how to tell, with real usage data, whether your team has outgrown it.
The Codex limit for Plus users is set per five-hour period, with weekly limits that may apply on top, and OpenAI pitches the $20 plan at "a few focused coding sessions each week." Before paying for Pro or Business seats, measure what your engineers actually use; a free tracker such as UsageAtlas gives you that data in an afternoon.
That is the short version. The rest is for anyone deciding what to put on the company card.
What Plus buys
OpenAI lists Plus at $20 a month. It covers Codex on the web, in the CLI, in the IDE extension and on iOS, plus cloud integrations such as automatic code review and Slack. When the allowance runs out, Plus users can buy ChatGPT credits and keep going without switching plans.
The allowance itself is not a fixed count. OpenAI publishes ranges of local messages per five-hour period and says plainly that these are estimates. The model, the size of the task, the context the session holds and whether work runs locally or in the cloud all move the number. Local and cloud work draw from one pool.
Where Plus starts to pinch
From a cost point of view there are three signals worth watching:
- The five-hour stop. An engineer pauses mid-afternoon to wait out the window. Annoying, but short.
- The weekly stop. A heavy Monday and Tuesday leaves nothing for Thursday. This one costs real time.
- Credit creep. The team keeps topping up with credits. At some point that spend passes the price of the next tier.
The options side by side
| Plan | Price on OpenAI's page | Five-hour limit | Fit |
|---|---|---|---|
| Plus | $20 a month | Yes, estimated per period | A few focused sessions a week |
| Pro | From $100 a month | None at present | Heavy individual use |
| Business | $20 per user a month, billed annually, 2+ users | Same estimates as Plus for Standard | Shared workspace, admin controls |
| API key | API pricing | Billed per token | CI and automation, no cloud features |
Weekly limits may apply across the board, and prices change, so confirm on OpenAI's pricing page before you buy.
Measure before you upgrade
Most teams guess. A better input is a week of real numbers per person: how often each engineer hit a window, which days ran hot, and what the work would have cost in tokens.
UsageAtlas is the tool we would hand an engineering lead for this. It is a free, open-source desktop app for Windows, macOS and Linux. It reads Codex usage through the Codex CLI each engineer is already signed in to and shows session and weekly windows, remaining capacity, plan and credits where reported, and a daily history with estimated cost. No account is needed, and it does not read prompts or responses, which makes it an easy ask of a team. An optional free account keeps history across machines.
Two caveats. Its cost figures are estimates, not your OpenAI bill, and provider data can lag. For the billing record, use OpenAI's own settings.
A simple rule
If nobody hits the weekly limit in a normal week, stay on Plus and let the occasional credit top-up absorb the peaks. If the same people hit it every week, compare their credit spend with Pro. If you need shared admin and data controls, Business is the reason to move, not the limit.
For a solo founder who opens Codex a few times a week, none of this is needed; /status in the CLI will tell you all you need.