# Google connects Gemini Spark to Chrome for logged-in web tasks > Source: > Published: 2026-07-30 21:14:59+00:00 Google began rolling out a [Gemini Spark integration with Chrome auto browse](https://x.com/geminiapp/status/2082923048362299629?s=46) on July 30th, giving its autonomous agent permission to navigate websites through a user's active browser session. The [Google Gemini (@GeminiApp)](https://x.com/geminiapp) account said Spark can handle tasks such as scheduling apartment viewings and filling in flight information. The integration connects two products Google introduced separately earlier this year. [Chrome auto browse](https://support.google.com/chrome/answer/16821166) arrived on January 28th as an agentic feature for completing multi-step web chores. [Gemini Spark](https://blog.google/innovation-and-ai/products/gemini-app/next-evolution-gemini-app/) followed at Google I/O on May 19th as a cloud-based agent designed to keep working after a user closes a laptop or locks a phone. Josh Woodward, who leads Google Labs, the Gemini app and AI Studio, introduced Spark in May. Google says Woodward previously helped co-found its Next Billion Users initiative and create Chromebooks. His group built Spark on Gemini 3.5 and Google's Antigravity agent harness, with connections to Workspace products including Gmail, Docs and Slides. The Chrome rollout delivers a capability Google previewed at I/O: letting Spark operate inside the browser rather than relying solely on cloud tools and remote browsing sessions. ### Spark gets access to the signed-in browser When a user approves a task, Spark can use local desktop Chrome and access the same websites the user is already signed into. Google says the agent can shop, compare products, make reservations, book travel and schedule appointments. It may also draw on connected apps, previous chats, location data and information supplied through Google's Personal Intelligence features. Spark can request permission to use Google Password Manager to sign into a website. Google says Password Manager does not expose the underlying password to Gemini, though Spark can use the credential to enter an authorized account. Users must review Spark's plan before a task begins and can stop the agent, take over the browser tab or resume the task later. Local browsing requires the computer and Chrome to remain awake. If the device becomes unavailable, Spark may shift the work to a remote browser. Tasks that require an interactive sign-in can pause until the user takes control. Access is initially limited to U.S. users aged 18 or older with Chrome set to English and a personal Google account subscribed to [Google AI Pro or Ultra](https://gemini.google/us/subscriptions/). Google AI Pro costs $19.99 a month and permits up to 20 multi-step Chrome tasks per day. Ultra starts at $99.99 a month and permits up to 200, according to Google's current help and pricing pages. Google says additional regions will follow. ### Google's security documentation spells out the tradeoff Giving an AI agent an authenticated browser session creates a broader security problem than adding a chatbot to a sidebar. Google's consumer documentation labels auto browse experimental, warns that the agent can make unintended purchases or take incorrect actions, and says users remain responsible for its activity. Google also warns that malicious instructions embedded in websites, emails or documents could attempt to make the agent reveal private information or send data to an outside service. Spark may share information such as names, contact details, files and preferences with third-party websites when needed to complete an approved task. The language is even sharper in Google's [Chrome Enterprise policy](https://chromeenterprise.google/policies/gemini-spark-settings/). Google tells administrators that enabling Spark with Chrome permits autonomous, multi-step actions through an active browser session and entails "significant security risks," including credential exposure, local-network ingress and the loss of context-aware security signals. The feature is off by default for managed users, and administrators can control whether Spark connects to Chrome. Google has been expanding Spark quickly since its May introduction. A [June 30th update](https://blog.google/innovation-and-ai/products/gemini-app/gemini-spark-updates-june-2026/) added a macOS beta, custom Model Context Protocol support and connections to services including Canva, Dropbox, Instacart, OpenTable and Zillow Rentals. Chrome now supplies the missing general-purpose interface: the websites and authenticated sessions where many tasks actually happen. That distribution gives Google an advantage over independent agentic-browser products. Google controls Chrome, the Google account, Workspace integrations and Password Manager, allowing Spark to act across services without asking users to adopt a separate browser. It also concentrates the risk inside one account and browser session, making Google's permission prompts, monitoring controls and resistance to prompt injection central to whether users trust Spark with consequential tasks.