Give your eve agent a browser

Your eve agent can now navigate the web like a human with agent-browser.

The @agent-browser/eve extension gives any eve agent a full set of browser tools: navigate pages, read content, click, fill forms, take screenshots, and inspect console and network activity. Everything runs inside the agent's sandbox.

Install the extension:


Then mount it under agent/extensions/:


Your agent gets namespaced tools like browser__navigate, browser__snapshot, browser__click, browser__fill, and browser__screenshot. Snapshot refs such as [ref=e12] become @e12 selectors, so the agent can inspect a page, then act on exactly what it saw.

The extension includes:

• 
  Sandbox isolation: Chromium and agent-browser run in the eve sandbox, not your app. Pre-install them in your sandbox template so sessions start warm, or let the extension install them on first use.
  
  
• 
  Domain allowlist: Restrict navigation and subresources with allowedDomains, and control output size, screenshot handling, proxy use, and session naming through extension options.
  
  
• 
  Credential protection: Cookie, storage, and saved-auth-state commands are never exposed to the model. Use eve extension overrides to disable tools, require approvals, or add app-specific guarded tools.
  

See a complete Next.js eve app with the extension mounted in the example, or read the documentation to get started.



Read more

Continue reading...