Browser Automation
The web is the world's largest API. Hermes Agent navigates it using real browsers, not brittle scrapers — enabling robust automation of any website a human can use.
What is Agent-Native Browser Automation?
Agent-native browser automation means an AI agent controls a real browser to interact with websites. Unlike traditional web scraping (which parses static HTML), agent-native approaches execute JavaScript, handle dynamic content, click buttons, fill forms, and make decisions based on visual rendering — just like a human user.
Capabilities
DOM Navigation
Click elements, fill forms, scroll pages, and extract structured data from rendered HTML.
Screenshot Analysis
Capture and analyze page screenshots for visual understanding, UI verification, and accessibility checking.
Network Interception
Monitor, modify, and mock HTTP requests and responses for testing and data extraction.
Session Management
Maintain cookies, localStorage, and authentication state across multiple browsing sessions.
Playwright Under the Hood
Hermes uses Playwright — the industry-standard browser automation library — with support for Chromium, Firefox, and WebKit. This gives Hermes access to the same tooling used by thousands of companies for end-to-end testing, but repurposed for autonomous agent behavior.
Learn More
Frequently Asked Questions
Does Hermes use a real browser or simulated one?
Hermes uses Playwright to control real browser instances (Chromium, Firefox, WebKit). This means it interacts with websites exactly as a human would — rendering JavaScript, handling cookies, and respecting rate limits.
Can Hermes handle JavaScript-heavy sites?
Yes. Because Hermes uses Playwright with full browser engines, it can execute JavaScript, wait for dynamic content to load, and interact with SPAs (Single Page Applications) just like a real user.
Is browser automation safe?
Hermes runs browser sessions in isolated sandboxes (Docker containers or VM environments). Sessions can be checkpointed and rolled back, ensuring that malicious or buggy web interactions do not affect the host system.
What about CAPTCHAs and bot detection?
Hermes respects robots.txt and terms of service. For legitimate automation tasks, it can use services like 2captcha or browser fingerprint randomization. The goal is ethical automation that websites permit.