What is Gemini Spark (now Task)? Google's personal AI agent, what it does, who gets it and its limits

Gemini Spark is Google’s answer to the always-on AI agent. Instead of answering one prompt at a time, it takes a goal, works on it in the background using your apps and the web, and can repeat the job on a schedule. This guide explains what Spark is, what it can reach, who can use it, what the limits are and how to turn it off, based on Google’s Gemini Apps Help pages and its announcement of skills. It is described as experimental, and we have not tested it.

Update, October 22, 2026: Google has renamed Gemini Spark to Task. You no longer switch to Spark from the sidebar. Instead, you start a task from the Chat and Task toggle next to the prompt box, in the same place you start a chat. Your existing tasks and schedules are in the sidebar. Google says it made the change because people wanted help finishing tasks. The steps below use the new names.

What Gemini Spark (now Task) is

Google describes Spark, now Task, as a personal AI agent that can automate complex workflows and manage schedules for ongoing tasks in Gemini Apps. You can now start tasks in the same place you start a chat. You tell Gemini your goal, and when you want it done, and Gemini works through it while you do something else. Google’s own examples are:

• Declutter your inbox: summarise or archive newsletters and unsubscribe from mailing lists.
• Get a custom news digest: go deep on stories you care about and follow how they evolve.
• Deep dive on topics: pull research formatted to fit your goal, with cited sources.

Spark can also handle web tasks through Chrome auto browse (desktop only for now), such as shopping, booking travel or making dinner reservations. It asks you to take over when it needs you to act in its browser. To turn a vague idea into a first task, Google suggests entering “set up”, “get started” or “interview me” in a Spark task.

What Spark can use

When you use Spark, Gemini draws on several sources, plus information from the websites it interacts with, including ones you are logged into:

• Skills and schedules, reusable instructions and automatic triggers.
• Connected Apps and other Google services, including custom connected apps.
• A remote browser and computer, where Gemini browses and runs code on your behalf.
• Personal Intelligence, Gemini’s use of your own context.
• Other Gemini features, and files you upload, add from Drive or attach as a Notebook.

Who can use it

Spark started as a first-access perk of Ultra, and Google’s US plan page still lists “Tasks in Gemini (in select countries)” among Ultra’s first-access features, as our Google AI Plus vs Pro vs Ultra guide notes. The help page lists AI Pro or Ultra, so Pro subscribers can use it too. Pro is $19.99 a month in the US and Ultra starts at $99.99.

How to create a task

According to Google’s help page, in the Gemini web app:

1. Go to gemini.google.com and click Task below the text box. On a Mac you can also use the Gemini app.
2. In the text box, describe the task you want Gemini to manage, including when to run it, for example around a specific time or after an event.
3. To use an existing skill, type @ and select it. To use a custom connected app, tell Gemini which one.
4. To add files, click Upload and tools, then upload files, add from Drive or add a Notebook.
5. Click Submit, and check back in case Gemini needs input. If it needs you to act in its browser, it asks you to take over.

Your tasks and schedules are listed in the sidebar, and each task thread has a work panel showing progress, files Gemini read or updated, schedules and permissions. Our guide to Gemini skills replacing Gems covers skills in detail.

Schedules

A schedule is the trigger that runs a task without you asking, and Google says it can be based on time intervals, like every Monday morning, or conditions, like an email arriving in your inbox. Google notes that schedules in Spark are different from scheduled actions in ordinary Gemini chats. Schedules don’t run while task mode is off or if your device is off, which Google lists in its turn-off notes.

Limits

Gemini Apps have compute-based usage limits that increase with higher Google AI subscriptions, and how much Spark consumes depends on how many requests you make and how complex they are. Google also sets a hard cap: up to 15 tasks can run at a given time, and a schedule won’t run if 15 are already running. Gemini tells you when you’re close to a limit.

Safety and privacy

Google says “Gemini’s ability to handle tasks is experimental” and stresses your supervision. It advises:

• Being careful about what you type directly in a task thread, and avoiding scheduling tasks you find sensitive.
• Monitoring tasks and taking over when Gemini asks, and stopping Gemini when needed.
• Knowing how Spark behaves with Gemini in Chrome.

You can delete remote browser, remote code execution and local browser data from the task’s work panel. If you turn task mode off (on the web, Settings, then Task settings, then Turn off under “Use task mode to hand off work”; in the Mac app, Gemini, then Settings, then Tasks), your remote browser data and remote code-execution data are deleted, schedules stop running and tasks are paused, but past tasks, threads and files that Gemini created are not deleted. Your Chrome browsing data isn’t erased. Turning task mode back on resumes paused tasks and schedules.

How Spark compares

Spark’s closest rivals are OpenAI’s new ChatGPT dots and Anthropic’s Claude Cowork. We set the three side by side in dots vs Gemini Tasks vs Claude Cowork. For the wider idea of software that acts for you, see AI agents explained and the Gemini profile.