Hark, the AI startup that secured $700 million in Series A funding in May, has unveiled Hark Handoff, a browser-use agent designed to efficiently handle online tasks without relying on official APIs.
What Hark Handoff Does
The company claims that Handoff can navigate websites like Target, Walmart, OpenTable, and LinkedIn — including those with no public APIs — by analyzing page structure and visual cues to determine where to click or what to type. This approach enables the agent to execute a wide range of tasks, from ordering food and booking travel to filing returns and conducting research.
Demo and Capabilities
In a demo, CEO Brett Adcock showed Handoff interpreting a natural language command to build a flower bouquet, accommodating vague preferences like "some of the florist's choice." However, the video only shows part of the process, leaving its real-world effectiveness open to question.
Under the Hood: A Different Model Architecture
For this release, Hark uses a post-trained model, with plans to pre-train later in 2026. The company says this iterative approach lets it rapidly refine its data pipeline and training infrastructure. Unlike LLMs that predict the next token, Handoff predicts the next action — such as a click or keyboard input at a specific location — making it more suited to task completion.
Competitive Landscape
Hark faces stiff competition from tech giants like Google, OpenAI, and Anthropic, as well as startups such as Browser Use, Polar, Strawberry, and Aside — all vying for dominance in the browser automation space.
Pricing and Availability
Hark asserts that Handoff is both faster and significantly cheaper than rival models like GPT 5.5 and Opus 4.8. The startup has opened a waitlist and plans to release the platform by the end of summer 2026.
via TechCrunch AI