H Company Launches Runner H Public Beta, Holo-1 Visual Model, and Tester H for Developer Workflows

Introducing Runner H: A New Era of Agentic AI Automation

H Company, based in Paris, has unveiled Runner H in public beta, a groundbreaking AI agent designed to automate complex, multi-step workflows seamlessly. Runner H acts as a digital coordinator that breaks down high-level goals into actionable tasks executed by specialized sub-agents such as Surfer H, a browser agent. This system enables automation of tasks like data extraction, spreadsheet compilation, Slack notifications, job application processes, CRM updates, and more.

How Runner H Works

Runner H addresses common developer frustrations with large language models by integrating memory, orchestration, and execution. Instead of delivering mere textual advice, it allocates tasks to mini-agents capable of planning, API calling, UI interaction, and tracking prior actions. This unified approach replaces fragmented workflows involving scripts, third-party tools, and manual checks with a single chat-based interface.

Key Features of Runner H

• Agent Orchestration: Initiate entire workflows from a single prompt, with Runner H coordinating AI agents to complete steps synchronously.
• App Integration: Seamlessly connects with Slack, Notion, Google Sheets, and various APIs for immediate automation within existing software ecosystems.
• Knowledge Uploads: Users can upload PDFs and data files that Runner H processes to enhance context and improve output accuracy.
• Autonomous Payments (Upcoming): Planned functionality to allow Runner H agents to make purchases or subscribe to services without human intervention.

Getting Started with Runner H

To try Runner H, visit the H Company website, sign up, and submit a prompt. For example:

Visit marktechpost.com and give me summarize of the 3 latest AI research news articles.

Runner H will execute the task, demonstrating its coordination of multiple sub-agents.

Holo-1: Open-Source Vision Models for Browser Automation

H Company also released Holo-1, a set of 3-billion- and 7-billion-parameter vision-language models optimized for UI localization and browser automation. Holo-1 achieves a 92.2% score on the WebVoyager benchmark and is efficient enough to run on a single GPU. Both the model weights and the WebClick dataset (1,639 scenarios) are available on Hugging Face, supporting academic and startup development of visual agents.

Tester H: Enhancing Quality Assurance with AI

Tester H, now in private beta, converts plain-English user stories into executable end-to-end tests. It simulates real user interactions by clicking buttons, opening modals, verifying text and layouts, and reporting discrepancies. Early adopters report reduced regression cycles and increased deployment confidence. Enterprise users can request beta access via H Company’s website.

The Future of Agentic AI

These three launches together represent a practical roadmap toward integrating multi-agent AI systems into everyday engineering workflows. H Company envisions 2025 as the year when agentic AI transitions from demos to essential productivity tools, changing how developers and businesses handle automation and quality assurance.