Runner H: Automating Web Page Execution Tasks through Natural Language Commands (Apply for Internal Testing)

General Introduction

Runner H is a company dedicated to developing cutting-edge action models designed to enhance worker productivity through advanced AI capabilities. Its flagship product, Runner H, is an advanced AI agent designed to help users automate complex, multi-step tasks and reduce repetitive and manual input. Through simple natural language commands, Runner H can design and execute web automation processes, dramatically improving the efficiency of developers and teams in web testing and process automation. The platform also provides a development environment called Studio, where users can create, run and extend automation tasks.Runner H's core technologies include Visual Language Modeling (VLM) and Large-scale Language Modeling (LLM), enabling it to understand and adapt to changes in the user interface and fix problems automatically.

Function List

• Task automation: Eliminate multi-step manual work by automating the design of web page automation processes through natural language commands.
• Decision support: Provide intelligent decision support to help users make smarter business decisions.
• Interface Adaptation: Automatically adapts to UI changes, reducing the complexity of maintaining selectors.
• self-healing function: Self-healing capability to automatically repair automation outages due to interface changes.
• Developer Platform: Provides Studio, a developer platform that supports automated processes for creating, reviewing, and editing.
• API integration: Support a wide range of automation needs by calling agents running in the cloud via APIs.

Using Help

Installation and use

1. Create automated processes::
   • After logging in, go to Developer Platform Studio.
   • In Studio, click the "Create New Process" button and enter a process name and description.
   • Use natural language commands to describe tasks that need to be automated, such as "Click the Add to Cart button".
2. Editing and Debugging::
   • After you create a process, you can view and edit each step of the process in Studio.
   • Using the debugging tools provided with the platform, run the process and check for errors.
   • If the process is interrupted, utilize the self-healing function to automatically repair or manually adjust the process steps.
3. Operation and Monitoring::
   • After editing and debugging, save the process and click the "Run" button.
   • During operation, you can monitor the execution of processes in real time and view logs and results.
4. API call::
   • For those who need to integrate with other systems, they can use the API provided by H Company.
   • Generate API keys on the platform and refer to the API documentation for integration calls.

Main function operation flow

Task automation

• Description of the task: In Studio, use natural language to describe the steps of a task that needs to be automated.
• Generation process: The system automatically generates the corresponding automated process based on the description.
• Debugging and Optimization: Run the process, check for errors and make optimization adjustments.

Decision support

• data entry: Enter the data to be analyzed and the system will automatically generate decision recommendations.
• View Report: System-generated decision reports can be viewed and downloaded from the platform.

Interface adaptation and self-healing

• auto-adaptation: The system automatically adapts to changes in the web interface, reducing the need for manual adjustments.
• self-healing function: In the event of a process interruption, the system automatically attempts to repair it to ensure that the process runs smoothly.

Featured Functions

• Visual Language Model (VLM): Runner H's VLM understands and adapts to changes in the user interface, automatically fixing problems and improving the stability of automated tasks.
• Large Language Model (LLM): Runner H's LLMs have basic programming and high-level decision-making capabilities that enable them to handle complex automation tasks.
• self-healing function: Runner H automatically fixes problems caused by user interface changes, reducing the time and cost of maintaining automated tasks.

Make AI Work for You with Runner H

Navigating the interface, reading and interpreting the documentation, clicking on the right components, and then repeating these actions. Currently, these tasks are performed by humans every day - sometimes all day.

It's different now.

Today, we're proud to introduce Studio, a platform designed for developers (and ultimately anyone) to easily create production-ready and highly reliable automated processes at scale.Studio is scalable and will continue to grow to support diverse automation needs and generalize beyond the web.

Our flagship agent, Runner H, will be available in private beta on the platform. With simple natural language commands, it automates the design of web automation processes, eliminating cumbersome multi-step manual operations and revolutionizing the way developers and teams approach web testing and process automation.

As a web proxy, Runner H offers state-of-the-art performance, outperforming the public benchmark WebVoyager on the Anthropic Computer Use (Learn more). Our in-house base model - which is smaller, specialized and less expensive (possibly by orders of magnitude) - can outperform large general-purpose models when providing support for agents.

We're excited to see Runner H leading the age of agents - systems that can reliably perform complex tasks. If you'd like to join us, pleaseSign up for the Private Beta Waiting ListThe

Studio and Runner H 0.1

Web developers often spend countless hours maintaining fragile selectors and fixing automation breakage issues caused by the ever-changing nature of modern web interfaces.

Runner H is our AI web agent powered by Studio that solves these pain points by automatically adapting to UI changes and self-healing by understanding natural language commands.

While our agents handle the complexity of writing and maintaining selectors behind the scenes, developers can focus on the semantic part of workflow and production, freeing up more time for higher-value development work.

The upcoming private beta includes:

• Calling APIs from ready-to-use and managed agents in the cloud
• Studio for creating automation, reviewing and editing the past and running in real time

In Studio, you can create reliable automated processes for complex workflows, such as end-to-end testing of e-commerce scenarios (from product discovery to order confirmation) and onboarding processes for financial services (pre-filled multi-step validation processes, document uploads and compliance checks).

An important step towards a grand vision: plan, observe, run and repeat

Compared to traditional methods such as screen automation, Runner H delivers production-level automation more reliably - efficiently, easily and with the ability to scale. This is due to the accomplishments of our engineering and research teams since the company was founded:

• We have trained the best Visual Language Model (VLM) on the market to predict the coordinates of a mouse click given a screenshot and a natural language command (e.g., "click on the Add to Cart button"), and its performance excels in Screenspot, a benchmark that focuses on user interface manipulation modeling.
• We've designed the best web proxy on the market for open-ended tasks, and its performance excels in WebVoyager.

To learn about the foundations of Runner H, more about our VLM and Large Language Model (LLM) seehere (literary)The

The Way Forward

Ultimately, we envision a future where you can interact with Runner H as naturally as you interact with your coworkers.

Next, we will continue to move forward in the following areas:

• Improve accuracy and cost efficiency through large-scale complex techniques including reinforcement learning and distillation
• Add debugging and teaching features to Studio so developers can train Runner H 0.1 to excel at specific tasks
• Promote the developer community through technical content, support and events
• Adherence to enterprise-grade security standards to ensure that Runner H 0.1 operates safely and reliably (more research to come)

This is our first step on the road to democratizing agency. At H, we believe that reducing complexity creates room for innovation. When we spend less time on menial tasks, we can focus on meaningful goals and creativity. And when we let AI do the work for us, we can get more important work done.

We look forward to exploring the possibilities. Join us in shaping the future of web automation.