Simulations allow your AI agents to explore and learn your application’s user interface through actual interaction. This hands-on exploration helps agents understand how your product works, making them significantly more effective at helping users.

​
What Are Simulations?

A simulation is an automated exploration session where your AI agent interacts with your website or application to learn its interface, workflows, and functionality. Simulations are powered by cloud browser automation (Browserbase) — the agent explores your actual live application in a real headless browser. As the agent explores, it builds a knowledge graph of what it learns: page structures, UI elements, navigation patterns, workflows, and feature relationships. This knowledge graph becomes part of the agent’s understanding and directly improves its ability to assist users.

​
Why Simulations Matter

Most AI support platforms rely on static documentation, which leads to generic responses and outdated information. Simulations solve this by giving agents real, first-hand experience with your product:
  • Real UI understanding — Agents see and interact with your actual interface
  • Current knowledge — Agents learn from the live version of your product
  • Specific guidance — Agents can provide exact steps based on real exploration
  • Dynamic learning — Agents adapt as your product evolves

​
What Agents Learn

Interface elements: Button locations, form fields, navigation menus, modal dialogs, error messages. User flows: Account creation, checkout processes, feature access, troubleshooting steps. Feature relationships: How different parts of your application connect and work together.

​
The Simulation Process

  1. Preparation — Agent receives instructions and application details
  2. Exploration — Agent opens your app in a cloud browser, navigates through interfaces, and interacts with elements
  3. Learning — Agent captures screenshots, records interaction sequences, and builds a knowledge graph
  4. Integration — New knowledge is integrated with existing information, ready to help real users

​
When to Run Simulations

  • After launching new features — Keep agent knowledge current
  • During initial setup — Give agents a thorough understanding of your product
  • When users report confusion — Explore areas where users struggle
  • On a regular schedule — Ensure knowledge stays up-to-date as your product evolves

​
Next Steps

  1. Create your first simulation — Set up and run a simulation
  2. Monitor simulation progress — Watch your agent learn in real-time
  3. Review simulation results — Verify what your agent learned
  4. Test in the playground — See how simulation learning improves agent performance