H Company frees the runner Heta Public alongside Holo-1 and test H for developers

The idea behind the agency's AI is that many small agents focused on tasks can cooperate to finish real work; However, this particular idea looked more like a promise than a product. Fortunately, the company H based in Paris wants to change this, announcing 3 major measures in the life of our vision of agentic AI, starting by putting their flagship run on the starting line.

Runner h by the H Company is now in a public beta version, inviting anyone to draw a single prompt and to look at a stunt of sub-agents filling calculation sheets, scratching the sites, making a penguin or even adjusting an invoice while you scan the results.

The Runner H version comes next to two other announcements: Open-Sourcing The Visual Model which guides the cousin of the Runner H browser, the surfer H, and the launch of a private test platform entitled Test H. Together, these three announcements sketch a practical roadmap to bring multi-agent systems into daily engineering.

What is the runner H?

According to the company H, Runner H is a cutting -edge AI agent which allows anyone automatizes complex, bulky tasks and in several stages without repetitive and manual entry.

In simple words, Runner H is a coordinator of your digital workload:

You provide a high -level goal, and the internal orchestration system breaks down this. This system intelligently attributes tasks to specialized sub-agents, including their parachuin agent, the H surfer and other connected applications. This allows the runner H to:

• Extract the data live, compile them in a spreadsheet and share this with your teammates via Slack.
• Look for relevant online job openings and carry out the request process.
• Update the customer relations management system (CRM) with new lead information and write personalized monitoring emails for this prospect.

Runner H's design responds to the two frustrations that developers mention most with important language models (LLMS): fragmented context and low execution.

Instead of returning a paragraph of advice, the runner H allocates the tasks to the mini-agents that can plan, call APIs, click on the user interface and keep track of what happened the last time. H The company calls the approach “memory + orchestration + execution. “In practice, this means that tasks that bounce between a script of Shell, Zapier and a human verifier can live in a single cat wire.

Runner H incorporates memory, task orchestration, execution capacities and deep connections with other software in a single interface.

Here are the key capacities:

• Agent orchestration: Users can launch complete workflows using an prompt. The runner then coordinates various AI agents to plan and perform the necessary steps in a synchronized manner.
• Integration of the application: It connects with popular tools like Slack, Concept, Google Sheets and various APIs, which allows it to operate directly in your existing software configuration for immediate automation.
• Knowledge downloads: Users can provide PDF, documents and data files. Runner H processes this information, transforming it into an accessible context to clarify its actions and generate more precise results.
• Autonomous payments (coming soon): H Company plans to allow the agents of the Horur H to make purchases from platforms such as Amazon and Shopify or to subscribe to software services, all without direct human intervention.

According to H Company, Runner H is designed to redefine the interaction with AI by creating a more intuitive and powerful way to manage digital tasks.

How to use Runner H:

Step 1: Visit the runner H Web page On the company's website H and click try now.

Step 2: Register to start. Enter your prompt and submit.

• Fast: Visit MarkTechpost.com and give me a summary of the last 3 research press articles on AI.

Step 3: The runner H will take a few minutes to finish the task, depending on the complexity of the task. You can see and follow all from the start to the end.

Holo-1: open weight for visual navigation

Automation of the browser means seeing the browser. To make this company reliable and cheaper, the company has published Holo1A family of 3 billion and $ 7 billion vision language action models.

Twinned with Surfer hHolo-1 brand 92.2% on the webvoyer reference for the location of the user interface while remaining small enough to operate on a single GPU. The weights and a scenario of 1,639 Web click The data set is now live FaceGive academics and startups of fresh equipment for training agents who can read and click.

Tester H: Close the QA space

If the AI ​​launches the code in seconds, quality insurance cannot rely on weekly manual test plans. Test H, now in private beta for companies, transforms user stories into English in kind into executable ends. ‘

The agents would cross the interface as a real customer and begin to click on the buttons, to open modals, to check the text and the layout and to report any deviation.

The first tests indicate that they have reduced regression cycles while increasing confidence in each push. Requests for membership of the beta version are open to the company H site.

Conclusion

The recent announcements of H Company, in particular the launch of the runner H in the public beta version, is the first Grand Open playground that points to a future where AI agents play a more active role in personal and professional productivity. The addition of an open source visual model like Holo-1 and a testing agent like testing H contributes a precious resource to the wider IA community, directly targeting the DevOps pain points.

H Company bets that 2025 will be the year when the agents cease to be a demo and begin to be part of the tool chain. For the moment, the introduction of more sophisticated agental AI capacities is an interesting development to look in in increasing technological history.