TL;DR

Dots (OpenAI), Muse (Meta) and Grok Bot (xAI) are new always-on AI agents that keep working in the cloud after you close the chat. Dots targets ongoing projects inside ChatGPT, Muse targets personal tasks and has a free tier, and Grok Bot gives teams named AI coworkers. All three launched between August and early October 2026 and are changing fast.

Key takeaways

An always-on agent runs on its own cloud computer and checks in only when it needs a decision.

Dots lives in ChatGPT and needs a paid Pro plan; Muse has a free tier; Grok Bot comes with SuperGrok or Cursor plans, per launch coverage.

The core loop behind all three is simple: plan, act with tools, pause for approval on risky steps.

You can build a toy version of that loop in 15 lines and run it free in CoderFile.

What is an always-on AI agent?

A normal chatbot stops when you close the window. An always-on agent runs in the cloud with its own computer and browser, takes a standing job ("watch my inbox for unpaid invoices"), and comes back when it finishes or needs a decision. In August–October 2026, xAI, Meta and OpenAI each shipped one: Grok Bot, Muse and Dots.

OpenAI Dots

Announced at OpenAI’s developer conference on September 29, 2026, a Dot is an agent inside ChatGPT that keeps making progress between conversations. It’s strongest at ongoing research, reports and publishing, and hands coding work to Codex. Launch coverage lists it as requiring a ChatGPT Pro plan.

Meta Muse

Muse launched September 8, 2026 as a personal agent: bookings, shopping, forms, reminders and long-term goals, in its own app or WhatsApp. It’s the easiest to try because it started with a free tier.

xAI Grok Bot

Grok Bot (in beta since August 11, 2026) gives you a team of named bots with jobs — sales, marketing, operations — that work inside your tools 24/7. It is included with SuperGrok and with Cursor plans, though early testers report usage limits run out quickly.

Side by side

	Dots	Muse	Grok Bot
Maker	OpenAI	Meta	xAI
Built for	Ongoing projects	Personal life	AI teammates
Lives in	ChatGPT	App / WhatsApp	Your work tools
Entry price (launch)	Paid Pro plan	Free tier	With SuperGrok / Cursor

How they work under the hood

Strip away the mascots and all three run the same loop: plan the next step, call a tool, look at the result, and pause before anything risky. Here’s that loop in miniature — run it, then change the job list.

Build a real one next

Ready for more? Our 50-line agent tutorial adds tool calling, and the LLM API guide shows how to plug in a real model. If your agent code breaks, see 7 common agent bugs.

Sources: launch coverage from Less Admin HQ, Adam Digital and AI for Mortals, checked October 14, 2026. Prices change — confirm on each maker’s site.

Frequently asked questions

Which of the three is free?

At launch, Meta Muse had a free tier, while Dots required a paid ChatGPT plan and Grok Bot was bundled with SuperGrok or Cursor subscriptions. Check each maker’s pricing page, as plans are changing quickly.

Are always-on agents safe to give my inbox or card?

Start with one small, low-risk job and read the permissions. Good agents pause before irreversible actions like sending email or paying — the same approval step shown in the code example.

Can I build my own always-on agent?

Yes. The core is a loop that plans steps, calls tools and stops for approval. Run the example in this article, then open it in CoderFile and ask the AI assistant to add real tools.