OpenAI Dot vs. Meta Muse: the always-on AI agent showdown

Meta launched its always-on AI agent Muse on September 8 with a free tier and a $20 Power plan. OpenAI shipped a competing agent called Dot inside ChatGPT three weeks later, gating individual access behind a Pro 100 plan at $100 per month — five times Meta's mid-tier price.

The gap sets the tone for a comparison that runs deeper than sticker price. Both agents run their own cloud computers, plug into outside services, and keep working after the user walks away. The two companies pitched the same promise in September: hand the AI a task, then stop checking every five minutes.

What does each agent actually do?

Muse runs on a dedicated "Secure VM" with its own browser, Meta's documentation states. The interface mimics a chat with another person. Users reach Muse through its apps, the web, or WhatsApp, then connect email and calendar to widen what the agent can handle.

In an Engadget hands-on, Muse ingested saved Instagram and Facebook posts and reorganized them into recipe and restaurant collections. Meta says Muse can shop, complete forms, plan travel, send emails, and monitor prices. The agent can also link with services like email and calendars once a user is ready to expand what it does.

Dot lives inside ChatGPT and runs on what OpenAI calls GPT-6 Astra. Each Dot has its own cloud computer, draws on connected plugins, retains relevant context, and runs scheduled or recurring jobs. OpenAI says Dots can manage ongoing research, work across connected applications, and handle recurring work.

OpenAI is closely integrating the agents with ChatGPT, ChatGPT Work, and Codex. Users can also grant a Dot access to a local computer for file retrieval and web browsing. OpenAI plans to support running multiple Dots per user at the same time.

How much autonomy will the agents actually take?

Both companies now frame their products as supervised, not autonomous. Meta states that Muse asks for permission before sensitive actions such as sending an email or making a purchase. OpenAI pairs Dots with permissions, "Custom Rules," and an "Auto-review" layer to screen what they can do. Both companies caution that their agents can make mistakes.

The first hard limits come from third parties. Amazon has already blocked Muse from shopping on its site — one telling restriction on the "let the AI buy things" pitch. Every merchant, bank, and SaaS dashboard now faces the same choice: allow agent transactions, or refuse at the door.

How much does each one cost?

Pricing is the starkest split. Muse ships with a free tier capped by usage, a Power tier at $20 per month, and a Maximum tier at $100 per month. OpenAI's lowest individual access tier, Pro 100, costs $100 per month.

Business Premium seats cost $100 per user per month on annual billing or $125 per user per month billed monthly. Enterprise access remains in beta.

The subscription is not the only cost. Both agents grow more capable the more accounts a user links. Every connected email, file folder, or browser session is another piece of personal territory handed over.

Meta is using price as a wedge to onboard individual users, offering a free entry tier and a $20 Power plan that undercuts any OpenAI offering. OpenAI is betting that enterprise contracts, Codex integration, and tighter ChatGPT hooks will justify its $100 entry fee.

What happens to the data the agents see?

Meta states that conversations and data within Muse's VM are not shared with Meta's advertising systems. Users can opt out of having their interactions used to improve Meta's AI models. Engadget's hands-on found the model-improvement toggle enabled by default, alongside a default permission setting that prompts for only some actions.

OpenAI states that activity from personal-plan Dots may be used to improve its models depending on the user's ChatGPT setting. Business and Enterprise data is not used for training by default. OpenAI adds that proactive Dot research stays read-only until a separate action passes normal permission checks.

Who can use them right now?

Availability splits sharply. Dots are rolling out to eligible Pro users outside the European Economic Area, Switzerland, and the UK. Business Premium users in supported ChatGPT regions gain entry. OpenAI says Dot usage will not count against usage limits unless the Dot operates inside ChatGPT Work or Codex.

Muse is currently available in the US and Canada only.

The two rollouts now overlap in the United States and Canada — and nowhere else. The regional carve-outs turn the comparison into a quiet policy test. Dots are blocked across the EU, UK, and Switzerland. Meta keeps Muse at home in North America.

Why does any of this matter?

The two September releases turn always-on agents from a research curiosity into a real consumer and enterprise product category. Meta and OpenAI shipped their first frontier-lab agents that do more than wait for the next prompt within the same month.

The architecture, pricing, and privacy decisions made this September will define the consumer AI-agent category for the next product cycle. OpenAI's regional exclusion list and Meta's North America holdout signal how seriously each lab treats data-residency and regulatory exposure.

Watch the blocklists next. Every merchant that refuses Muse today is a merchant deciding whether to let any always-on agent transact tomorrow. That slow, public fight — already starting with Amazon's ban on Muse shopping — will shape which tasks these agents can actually perform, and which ones will keep stopping at the checkout button.