OpenAI's new Dots agents run 24/7 without consuming usage allowances during the launch phase, but tasks delegated to products such as Codex will count against metered limits, influencing developer infrastructure decisions.
- Dots run continuously without using plan allowances unless calling metered services.
- Delegation to Codex or other paid APIs triggers consumption of usage quota.
- Future paid upgrades will offer higher speed and bandwidth for Dots.
Infrastructure signal
OpenAI’s introduction of always-on Dots represents a significant shift in cloud infrastructure cost models by embedding continuous agent operation within user plans at no extra charge during the initial launch period. This approach absorbs operational costs traditionally billed separately and potentially reduces barriers for constant background agents.
However, when these Dots delegate tasks to metered services like Codex, such activity is billed against usage quotas tied to those services. This means that while baseline cloud consumption might appear free or nearly free, cost can scale unpredictably based on agent delegation patterns. Operators should anticipate variability in billing tied closely to workflow design.
Developer impact
Developers implementing Dots must carefully architect workflows to balance between baseline free usage and metered API calls. Since a single Dot can internally route tasks either within free allowances or to paid services like Codex, decisions on task delegation will directly affect both cost and performance.
Additionally, developers should monitor usage carefully, as OpenAI’s current terms do not guarantee alerts before transitioning from free to metered usage, increasing the risk of unexpected charges. The promise of future paid enhancements to increase Dot speed and bandwidth also indicates evolving developer options for scaling agent capabilities.
What teams should watch
Engineering and product teams should track the evolving usage policies following the launch period because OpenAI's initial free usage window is limited to the first month, after which clear per-plan terms for included Dot usage will be published. This affects budgeting and capacity planning for cloud infrastructure.
Teams should also evaluate alternative solutions, such as AWS’s open source agents which advertise lower operational costs, to optimize cost-efficiency. Observability tools that correlate agent invocation paths with billing impacts will be critical to avoid surprises as agents autonomously choose between free and metered execution paths.