Gemini Spark: Google's 24/7 AI Agent Joins Your Workflow

At Google I/O 2026 on May 19, Google unveiled Gemini Spark, a 24/7 cloud-hosted personal AI agent that runs continuously, even when your computer is off. Spark goes beyond conversational AI: it manages tasks inside Gmail, Google Docs, and the broader Workspace suite, connects to third-party tools via the Model Context Protocol, and requires user approval before taking any high-stakes action such as sending an email or scheduling a calendar event.

What Happened

Google announced Gemini Spark at the I/O developer conference and described it as turning Gemini from "an assistant that answers questions into an active partner that does real work on your behalf." Spark runs in the cloud and is not tied to any single device session. Assign it a task before you leave your desk, and Spark continues working after you close your laptop.

The underlying architecture combines Gemini 3.5 with what Google calls an "agentic harness from Google Antigravity," the same team behind the Antigravity CLI announced earlier at I/O. MCP (Model Context Protocol) connections expand Spark's reach to third-party services. At launch, confirmed MCP integrations include Canva, OpenTable, and Instacart, with more on the roadmap.

Gemini Spark is rolling out to trusted testers this week. US Google AI Ultra subscribers ($100 per month) will receive access next week. A broader rollout timeline has not been confirmed.

Why It Matters for Creators

Every AI assistant on the market today is session-based: you open the app, give it instructions, and it works until you close the window. Spark changes that model. It operates continuously, which means it can monitor incoming emails for project updates, surface threads that need a response, and draft replies, all without you initiating a new conversation.

For creative professionals managing multiple client relationships, this addresses a real friction point. A freelance designer working with three clients simultaneously does not need to check three email threads every morning to stay current. Spark can triage that work in the background and surface only what requires a human decision.

The Canva integration via MCP is notable. 9to5Google confirmed Canva as a launch-day MCP partner alongside OpenTable and Instacart. When Google's own design tool (Google Pics, also announced at I/O 2026) and a leading third-party design platform are both accessible through the same agent, the possibility of cross-platform asset management becomes real. Spark could theoretically move a generated image from Pics into a Canva layout without the user switching between apps.

Core Features

• 24/7 cloud operation: Spark runs in Google's cloud infrastructure continuously. It does not require an open browser tab or an active device session.
• Gmail and Workspace integration: Out-of-box access to Gmail, Google Docs, Google Slides, Google Calendar, and other Workspace apps. No setup or permissions configuration required beyond the initial subscription.
• MCP third-party connections: Spark connects to Canva, OpenTable, Instacart, and additional third-party services via the Model Context Protocol. Google plans to expand this list over the summer.
• User-approval gate: Before taking sensitive actions, such as sending a composed email or adding a calendar event, Spark presents the action for user review and waits for approval. It does not act unilaterally on high-stakes tasks.
• Sub-agents: A summer update will let users create custom sub-agents, smaller specialized agents that Spark can delegate tasks to.
• Local browser control: Later in summer 2026, Spark will gain the ability to operate your local browser, enabling web-based tasks beyond the Workspace and MCP integrations already available at launch.
• Text and email access: A summer update will add the ability to reach Spark directly via text message or email, so you can assign tasks from any device without opening the Gemini app.
• Android Halo support: Spark will be supported on Android Halo later in 2026.

Gemini Spark vs. Microsoft Copilot vs. Claude

How to Use Gemini Spark for a Creative Workflow

Once Spark reaches your account, here is a practical setup for a freelance designer or content creator managing client projects:

1. Brief Spark on active projects: Open the Gemini app and tell Spark about your ongoing projects. Name the clients, the deliverables in progress, and which Gmail labels or threads are relevant. Spark indexes this context and monitors those threads continuously.
2. Let Spark triage your inbox: Spark monitors incoming Gmail for messages related to your briefed projects. Each morning, instead of scanning a full inbox, you review Spark's summary of what needs attention and what it has already drafted a response to.
3. Approve or edit before sending: When Spark drafts a client reply or schedules a follow-up, it surfaces the action for your review before sending. You edit, approve, or reject the action in one tap.
4. Use MCP to move assets: Connect Spark to Canva via MCP. When a client requests a revision to a design, Spark can open the relevant Canva file, surface the current version, and prepare a comment thread response from the email context.
5. Text Spark from your phone (summer): Once the text-access feature ships, assign tasks to Spark via SMS even when you do not have the Gemini app open. Spark handles the task and reports back.

Availability and Pricing

Gemini Spark is available to trusted testers starting the week of May 19, 2026. US subscribers on Google AI Ultra ($100 per month) receive access the following week. Google has not confirmed availability for Google AI Pro ($19.99 per month) subscribers or a free tier.

Planned summer additions include: text and email access to Spark, custom sub-agent creation, and local browser control. Android Halo support follows later in 2026.

The $100 per month price point places Spark alongside Google Pics and Gmail Live as features targeting the premium tier of Google's subscription stack. If you are already a Google AI Ultra subscriber, all three arrive with no additional cost.

Frequently Asked Questions

What is Gemini Spark?

Gemini Spark is Google's 24/7 personal AI agent, announced at Google I/O 2026. It runs continuously in the cloud, handles tasks inside Gmail and Google Workspace, and connects to third-party apps like Canva via the Model Context Protocol. It requires user approval before taking high-stakes actions.

How is Gemini Spark different from regular Gemini?

Regular Gemini is a session-based conversational assistant: you open it, ask a question, and it responds. Gemini Spark operates continuously in the background, monitors your email, completes multi-step tasks across apps, and works even when your device is off.

Does Gemini Spark send emails without my approval?

No. Spark requires explicit user approval before sending any email, scheduling any calendar event, or taking any other sensitive action. It drafts and presents the action for review; you approve or reject it before it executes.

What apps does Gemini Spark connect to?

At launch, Spark has built-in integrations with Gmail, Google Docs, Google Slides, Google Calendar, and Google Workspace broadly. MCP third-party connections include Canva, OpenTable, and Instacart. Additional integrations are planned for summer 2026.

When is Gemini Spark available?

Trusted testers receive access the week of May 19, 2026. US Google AI Ultra subscribers ($100/month) get access the following week. Broader availability, including for AI Pro subscribers, has not been confirmed.

How much does Gemini Spark cost?

Spark is included in Google AI Ultra at $100 per month. There is no confirmed lower-tier or standalone pricing.

Can Gemini Spark control my browser?

Browser control is planned for summer 2026 as an add-on capability. At launch, Spark operates through app integrations and MCP connections rather than direct browser automation. This will change once the browser control feature ships.