Runner H AI Agent
H Company · Productivity · Business
Runner H AI Agent is an AI web agent built by H Company that carries out browser tasks from plain-language instructions. Instead of writing brittle selectors or scripting every click, you describe what you want done and the agent navigates pages, fills forms, and follows through on multi-step jobs. It grew out of H Company's Studio platform and now runs alongside the broader Holo family of computer-use agents, which the company pitches for automation, testing, and research work.
About Runner H AI Agent
What Is Runner H AI Agent
Runner H AI Agent is a browser automation agent from H Company, a Paris-based AI lab. You give it a task in everyday language, and it works through a website the way a person would: reading the page, deciding which element to click, and moving to the next step. H Company introduced Runner H on June 3, 2025 as the flagship agent of its Studio platform. The pitch is simple. Developers and teams spend hours maintaining automation scripts that break whenever a site changes its layout. Runner H aims to take that job off their hands.
The core selling point is what H Company calls self-healing behavior. Traditional screen automation depends on fixed coordinates or CSS selectors, so a redesign can silently kill a workflow. Runner H instead interprets the page and adapts. The company says that lets it hold up when interfaces shift. Handy. On the public WebVoyager benchmark, H Company reported that Runner H outperformed Anthropic's Computer Use. That's the vendor's own comparison. Treat it as a claim, not a neutral result.
The biggest limitation is access. Runner H launched in private beta and the platform has since moved toward the Holo models and managed Computer-use Agents, so the exact packaging keeps changing. Check the official site before you commit. If you're evaluating it today, don't assume the original beta terms still apply.
Getting Started
- Head to the H Company website and go to the agent or Holo models section.
- Sign up for access, or ask about the private beta if you're an early evaluator.
- Open the Studio, then describe your first task in plain language. Something like finding a product and adding it to a cart works well.
- Watch the run in the interface to see which steps the agent takes and where it hesitates.
- Connect through the API, CLI, or MCP server once you're ready to move from a test run to a real workflow.
Product Information
A quick look at Runner H AI Agent's pricing, supported platforms, and performance.
Best for
The users, tasks, and scenarios where this tool fits best.
Users
- Developers building web automation
- QA teams testing live sites
- Operations and back-office staff
Tasks
- End-to-end e-commerce checks
- Form-heavy onboarding
- Web testing and regression runs
Scenarios
- A retailer wants to confirm the checkout flow still works after a redesign, without rewriting test scripts.
- A fintech team needs to verify that an onboarding form accepts the right documents across several account types.
- An analyst wants a competitor's pricing page checked on a schedule and the results collected in one place.
Key features
Natural Language Instructions
You describe the task the way you'd explain it to a colleague, and Runner H turns it into a sequence of browser actions. There's no selector syntax to learn up front. That lowers the entry bar for people who aren't developers. The agent reads the page to figure out what to do, instead of relying on hardcoded paths.
Self-Healing Automation
Sites change, and scripted automation usually breaks when they do. Runner H re-interprets the page on each run. So a moved button or renamed field doesn't automatically kill the workflow. That's the whole idea. H Company frames this as the main reason it holds up better than traditional screen automation on long-lived tasks.
Vision Language Models Under the Hood
The agent runs on H Company's in-house foundation models rather than a general-purpose model wrapped in a control loop. These are built specifically for clicking around a screen. H Company says the smaller, specialized models can beat larger generalist ones at click prediction, and that they cost less to run at scale. That matters if you're executing thousands of runs a month. Costs add up fast.
Studio for Building and Reviewing Runs
The Studio is where you create automations and review past and live sessions. You can watch what the agent decided at each step. That helps when a run goes sideways and you need to see why. It's aimed at turning one-off tasks into repeatable pipelines.
API, CLI, and MCP Access
You can call managed agents through a REST API, run them from a command line, or connect them through an MCP server. SDKs exist for TypeScript and Python. H Company says most users get a first agent running in under half an hour. That fits teams that want to embed the agent in their own tooling rather than work inside a web UI.
Composable Agents and Skills
Agents can be configured once and reused by name, and you can combine reusable skills or hand work to specialist subagents. Instead of cramming everything into one prompt, you build smaller pieces. Then let the system route between them. Every run stays visible, so you can trace what powered a given result.
Managed Infrastructure
There's no runtime to provision or model to maintain on your side. H Company runs the agents in the cloud and handles the execution layer. No servers to babysit. That's the difference between a demo and something you can leave running unattended.
Pros and cons
Pros
- Plain-language task setup removes the need to write and maintain selectors.
- Self-healing behavior means common site redesigns don't automatically break a workflow.
- In-house vision models are pitched as cheaper per run than large generalist models.
- API, CLI, and MCP options make it workable inside an existing developer stack.
Cons
- Pricing isn't published, so you can't estimate cost without contacting H Company directly.
- The benchmark advantage over Anthropic Computer Use comes from H Company's own testing, not an independent source.
- The product line has shifted toward Holo models and managed Computer-use Agents, so early beta details may no longer describe what's on offer.
Frequently asked questions
It carries out browser tasks from natural-language instructions, so it can navigate sites, fill forms, click through multi-step flows, and gather information across pages. Typical jobs include e-commerce checks, testing, and repetitive back-office work. Think of it as a tireless pair of hands.
Related content
Explore related tools, skills, and articles for Runner H AI Agent.
Runner H AI Agent Alternatives
Swms
Swms AI · Writing · Productivity · BusinessSwms is an AI safety compliance tool that turns a plain description of a job into ready-to-use safety documents. Describe your project, trade, and known hazards, and it writes a safe work method statement, job hazard analysis, safe work procedure, RAMS, or safety data sheet in seconds. It also ships with Oscar, a chat assistant that answers workplace safety questions in any language. Teams in construction, mining, transport, and warehousing use it to cut the paperwork that usually eats into a workday.
Wordly AI Translation
Wordly · Voice & Language · ProductivityWordly AI Translation is a real-time AI translation and captioning platform built for meetings, conferences, and events. It delivers live translation, captions, transcripts, and summaries in more than 60 languages, and attendees join by scanning a QR code or opening a link instead of using dedicated headsets. The platform works with Zoom, Microsoft Teams, Google Meet, and Webex, and it's designed for organizations that want multilingual access without hiring human interpreters for every session. Simple as that.
Rows
Rows (Superhuman) · Productivity · BusinessRows is an AI spreadsheet and data analysis tool that imports live data from more than 50 sources, then lets you work with it using plain language instead of SQL queries or nested formulas. You can pull numbers out of PDFs, connect ad platforms, databases, and bank accounts, and ask the AI to build reports, merge datasets, or set up models that recalculate themselves. It runs in the browser, works like a spreadsheet, and now sits under the Superhuman umbrella.