OpenAI’s new Dots can work around the clock without drawing down a user’s normal usage allowance during the launch period, but users might be surprised how that changes once an agent hands work to another OpenAI product such as Codex.Thibault “Tibo” Sottiaux, OpenAI’s head of core products and platform, explained how Dots usage works in a post on X. His launch announcement caused a Community Note to question whether always-on agents could realistically run inside existing subscriptions without blowing past usage limits.“I got community noted, but the note is wrong,” Sottiaux wrote. He continued in his post saying that a user’s primary Dot is included in the plan, is available 24/7, and uses none of the plan’s allowance when it does work directly. If the user asks that Dot to create a task in Codex, however, the Codex task draws usage as usual.If the user asks that Dot to create a task in Codex, however, the Codex task draws usage as usual.Included usage has an expirationSottiaux framed that arrangement as permanent, writing that the baseline functionality “will always just be included in your plan.” OpenAI’s published terms are narrower. The company says Dots usage won’t count toward eligible plan allowances for the first month after launch and that it will publish per-plan usage terms afterward. Developers planning workloads around Dots therefore don’t yet know what the included allowance will look like once that window closes.Sottiaux framed that arrangement as permanent, writing that the baseline functionality “will always just be included in your plan.”Codex tasks hit the meterSottiaux said OpenAI expects “a few million dots online within days.” Running that many agents continuously costs OpenAI money even when they never call another product. For the launch period, the company is absorbing that baseline activity into the subscription for each user’s primary Dot.Work a Dot starts in Codex or ChatGPT Work is treated differently and counts against those products’ limits. A Dot can sit online all day without touching the user’s allowance, then start spending it the first time it opens a Codex task. That allowance is also getting smaller, since OpenAI halved the included usage on its $200 Pro plan on the same day it launched Dots and added a $500 tier above it.Because a Dot can decide where to send work without anyone reviewing each step, where that work runs can directly affect what the user pays. Developers building around Dots will have to account for that the same way they account for model choice or token budgets.Sottiaux said OpenAI expects “a few million dots online within days.” Paid bandwidth is comingSottiaux also previewed a paid layer on top of the included Dot. “In the future you will be able to increase the speed and allow your dot to have more bandwidth,” he wrote, adding that the extra capacity will cost money while the baseline stays in the plan.“In the future you will be able to increase the speed and allow your dot to have more bandwidth.”Routing decides who paysA Dot can often finish the same job in several ways, whether it does the work itself, calls an included tool, or sends the task to a metered product like Codex, and the route it picks could determine whether the user is charged at all. Codex isn’t the only place delegated coding work can land, either, with AWS now pitching an open source agent that it says costs 45% less to run than Claude Code or Codex.Sottiaux didn’t say whether users will be warned before a Dot moves from included work into metered usage, which means users may not realize they’ve burned through their allowance until they check what’s left.The post OpenAI’s always-on agents are free, until one specific thing happens appeared first on The New Stack.