What H Company does
H Company (hcompany.ai) is a Paris-based AI research lab and product company building “computer-use” (GUI) action models and enterprise agents designed to operate inside real digital environments rather than only generating text. The company positions its core wedge as end-to-end agent stacks for the “autonomous enterprise” where agents must plan, execute multi-step tasks, and adapt when UI screens or states change.
More
Its product direction emphasizes (1) models purpose-built for GUI actions, (2) agent tooling to reliably execute those actions, and (3) an enterprise platform layer that can run agents either through APIs/managed infrastructure or on self-hosted/quantized checkpoints. For enterprise adoption, H Company highlights that it can automate workflows across legacy systems and environments that lack clean APIs by using “computer use” when necessary. Its enterprise offering describes orchestration across ecosystems, including legacy/mainframe/desktop environments, where agents switch to UI-based execution for steps that require it. The company also markets managed agent capabilities (computer-use agents) and a suite of supporting development/testing tools (e.g., its agent testing platform and model/agent interfaces for integrating into existing stacks). As of 2026-08-31, H Company is actively publishing model releases (e.g., Holo3.1, Holotron-12B) and engineering infrastructure posts (e.g., SkyPilot-based training operations) that indicate an emphasis on high-throughput, production-focused agent deployment. Business model (as described publicly): H sells enterprise agent capabilities and related software interfaces. It presents H Platform (including deployment, monitoring, and evaluation) as a place to configure/deploy/operate computer-use agents at scale, including via AWS Marketplace listings; it also offers model/agent access through API-style interfaces and developer tooling. Strategic position (based on available public information): the company’s current differentiation is “computer-use-first” agent modeling plus production-minded infrastructure and tooling (testing, managed execution, observability/traceability messaging). This approach targets enterprises where traditional automation and API-only agent designs struggle due to heterogeneous UI-based workflows and integration constraints.