OpenAI Dots ships always-on agents, and halves the $200 Pro limits
The 24/7 agents run on GPT-6 Astra across 4,000+ apps — but there's no public API yet, Pro excludes the EU/UK at launch, and Codex quota drops 20x to 10x.

Copy markdown
Meet Dots: ChatGPT's always-on agents
OpenAI's Dots run on GPT-6 Astra on their own cloud computers, chasing goals while you're away and keeping context between conversations. One named Dot comes with eligible plans, reachable from ChatGPT, Codex, Slack and Teams (SMS soon), and able to act across 4,000+ connected apps.
The catch: $200 Pro limits halved the same day
On launch day OpenAI cut Pro's Codex allowance from 20x to 10x and its weekly GPT-6 message cap from 200 to 100. The old allowance now sits behind a new $500/month tier — so a 'free' Dot arrives right next to a quiet 2x squeeze for heavy users.
Who gets one, who's locked out
Pro and Business Premium get Dots first, and chatting with your Dot doesn't burn ChatGPT limits (deep work inside Codex/Work still does). Pro excludes the EEA, Switzerland and UK at launch, and Enterprise is beta — off by default until an admin turns it on.
No Dots API yet — what you can build
As of Sept 30 there were zero public Dots API routes and no published model ID, so the always-on agent environment is app-only for now. You can still call the underlying gpt-6-astra via API, but you'd have to rebuild the persistence, scheduling and app connectors yourself.