H’s agents are designed to think, act, and reason like humans. They accomplish both simple and complicated tasks, defining objectives based on prompts, coming up with requirements and strategies to achieve those objectives, and then meeting them head on. The Agent Platform enables you to make the most of our state-of-the-art agents by integrating them directly into your apps and running them. This guide introduces our agents and shows you how to activate them.
Agents
Here’s a look at the currently agents we provide for you to learn about and use:
Surfer-H
Our flagship agent, Surfer-H, is designed to take action in the real world. It can see what’s on the screen, make decisions, and interact with a UI by clicking buttons and navigating from page to page. In doing so, it reliably fulfills tasks that previously required human oversight, from booking flights to researching online, and more. Surfer-H is advanced enough that it knows when it has completed what it set out to do, and can adjust, re-assess, and try again when it hasn’t. We’ve open sourced Surfer-H to share its power and groundbreaking abilities with the world.
Tester-H
Tester-H is a QA automation platform that helps product owners, developers, and QA teams build better websites and products by simulating user interactions and verifying expected outcomes. It automatically compares intended user flows with actual results to uncover successful journeys, bugs, and usability issues. By highlighting friction points and unexpected behavior, Tester H delivers valuable insights into how users experience a website.
Comparison chart
Here’s a table of our current agents and how they compare with one another:
| Agents | Description | Best for | Example use cases | Powered by |
|---|
Surfer-H | An AI agent designed to take action in the real world by seeing what’s on the screen, making decisions, and interacting with UIs. | Executing and automating real-world tasks. | Booking flights, online research, form submissions, end-to-end process automation. | Holo, our open, cost-effective VLMs designed to bridge the gap between visual perception and language understanding — enabling agents to interpret and act within web environments |
Tester-H | A QA automation platform that simulates user interactions, verifies outcomes, and identifies bugs and usability issues. | Ensuring product quality and usability in digital products. | UI/UX testing, bug detection, user journey validation. | Automated testing frameworks, behavior-driven testing, machine learning for anomaly detection. |
SDK instructions
Our lightweight Agent Platform JavaScript SDK lets you list or select agents and more.
List available agents
Use the listAgents() command to retrieve all currently accessible agents.
const agents = await agent.listAgents();
console.log('Available agents:', agents);
Get the best agent for web tasks
Use getBestAgent('web') command to select the most suitable agent for a specific action.
const bestAgent = await agent.getBestAgent('web');
console.log('Best agent for web tasks:', bestAgent);
Execute a task with a specific agent
Use the agent.run() command to execute your tasks.
const task = await agent.run(
'Search for "TypeScript tutorials" on Google and get the top 5 results'
);
Make sure to provide a task description. This enables the SDK to select the best possible agent based on the scope and nature of the task you’ve described.
.png)
