H Company

H Company is an AI startup that has raised $220 million to develop autonomous AI agents designed to execute complex tasks across digital workflows. Rather than simply providing answers like traditional chatbots, H Company's philosophy is "Chatbots give answers; Runner H executes"—representing a shift from conversational AI to what they call "action models" and "execution intelligence".

H Company has launched three main AI agents. Runner H is the flagship orchestration platform that enables users to accomplish complex, multi-step objectives through a single natural language prompt. It coordinates multiple specialist agents to plan and execute tasks across your connected applications. Surfer H is a web-native agent powered by Holo-1, H Company's proprietary visual-language model, designed to understand and act on web interfaces autonomously by combining visual and planning capabilities. Surfer H has achieved 92.2% accuracy on the WebVoyager Benchmark. Tester H is an enterprise solution for automated software testing that converts natural language test scenarios into executable automated testing protocols, supporting both web and mobile applications.

Runner H uses a "fleet" of smaller, specialized agents that work together rather than relying on one large general-purpose model. When you provide a single natural language prompt, Runner H spins up this fleet of specialist agents that plan tasks, call your apps (like Slack, Notion, Google Workspace, and Zapier), and can dispatch Surfer H to collect live web data before delivering finished work. The system can handle visual interpretation of web pages, automatically adapting to UI changes without requiring pre-defined APIs.

Runner H can execute a wide range of automations including filing out forms and multi-step verification processes, booking trips and scheduling, e-commerce workflows from product discovery to order confirmation, financial services onboarding with document uploads and compliance checks, robotic process automation (RPA), quality assurance and testing, business process outsourcing tasks, searching for specific information across websites, and working with your uploaded documents and data files to provide context.

Runner H is currently in public beta. H Company has been quietly working with customers in areas like e-commerce, banking, insurance, and outsourcing to refine the product based on real-world feedback.

During its current public beta, Runner H is free.

H Company claims that its compact model outperforms competitors significantly. Runner H's performance exceeds Anthropic's "Computer Use" by 29% based on WebVoyager benchmarks, as well as models from Mistral and Meta. This is achieved using a proprietary compact language model based on just 2 billion parameters—potentially orders of magnitude cheaper than large generalist models while delivering superior performance for agent tasks.

Holo-1 is H Company's proprietary visual-language model that powers Surfer H. The company open-sourced Holo-1 under Apache 2.0 licensing in partnership with Hugging Face, Nvidia, and AWS, enabling developers and enterprises to build production-ready custom automation solutions using visual AI capabilities.

Runner H can integrate with major productivity and business tools including Google Workspace, Slack, Notion, Sheets, Zapier, and APIs. Users can also upload PDFs, documents, and data files to provide context memory.

H Company's roadmap includes improving accuracy and cost-efficiency through reinforcement learning, adding debugging and teaching capabilities to allow users to correct agents, fostering a developer community around their APIs, and upholding enterprise-grade security and reliability. The company also plans to enable independent purchasing capabilities where Runner H can buy items from online stores like Amazon and Shopify or subscribe to software services without direct human involvement.