Hark, a startup that raised $700 million in funding in May 2026, previewed a tool called Handoff on August 5, 2026. The company has opened a waitlist and plans to release the platform by the end of summer 2026. (TechCrunch)

Handoff is designed to navigate websites like Target, Walmart, OpenTable, and LinkedIn and complete tasks on the user's behalf. Hark, led by CEO Brett Adcock, claims the agent can interact with these sites without needing each platform to build a special connection for it.

Most AI tools today work by predicting the next word in a sentence. Hark's model works differently: it predicts the next action, like clicking a button or filling in a form. For the current release, Hark adapted an existing model to do this. The company plans to build a new model from scratch later in 2026.

Hark's privacy policy, published March 5, 2026, describes a feature called the "Hark Browser Operator" that, when turned on by the user, gives Hark access to interact with web pages in the user's own browser. (hark.com/privacy-policy) This is how Handoff works on sites that were not designed to connect with outside tools.

The appeal of a tool like this is easy to understand. A huge number of everyday tasks, from buying groceries to booking a restaurant, still happen through websites built for people to click through by hand. If an AI agent could reliably do that for you, it would open up automation far beyond the services that offer official connections for developers.

The technical challenge is much harder to solve. Websites change constantly, and many are designed to block automated tools on purpose. Hark's approach of predicting actions rather than words suggests a system built specifically for navigating websites, rather than a language model with browser features added on. But the company has not yet released test results or outside evaluations to back up the approach.

Hark's plan to build a new model from scratch later in 2026 signals the company is still early in development. Building a model from the ground up requires significantly more computing power and data than adapting an existing one.

The $700 million funding round places Hark among the most heavily funded AI startups, and that level of investment is consistent with a plan that includes building a model from scratch. Whether that money is enough depends on details Hark has not shared: where the training data comes from, how large the model will be, and what it costs to run in real time.

Privacy is a natural concern with a product that needs access to your browser. The Hark Browser Operator is opt-in, meaning the user must turn it on. But the access it grants is broad by definition. A browser can hold your login information, payment details, browsing history, and data from every open tab. The privacy policy's wording is narrow, but what it allows in practice is wide-ranging. Users will need to weigh whether the convenience of having an AI handle browser tasks is worth granting a company this level of access.

Hark has not yet published technical documentation, model details, or independent security audits for Handoff. The waitlist and end-of-summer release window suggest a gradual rollout rather than wide availability right away.

The broader question is whether this kind of AI can become reliable enough across the open web to be useful in everyday life, not just in demos. Hark's preview is an early step toward answering that question, and the end-of-summer release will give a clearer picture.