Beginner · Systems

Agentic Browsers

Visual diagram · (in preparation) · Math · (in preparation) · Worked example · 3 difficulty levels.

TL;DR. Web browsers with a built-in AI agent that can read pages, fill forms, and complete multi-step tasks on the user's behalf.

Technical Definition

Web browsers with a built-in AI agent that can read pages, fill forms, and complete multi-step tasks on the user's behalf.

How it works

Agentic browsers moved the AI assistant from a sidebar that summarizes pages to an actor that operates the web: 'book the cheapest flight Tuesday morning' becomes a sequence of clicks, form fills, and confirmations the agent performs while the user watches or walks away. The 2026 generation added permission scopes per site, visible action logs, and confirmation gates for purchases. The dominant risks are prompt injection from malicious page content — the page is both the environment and a potential attacker — and over-broad credentials, which is why mature designs isolate the agent's session and require explicit approval for irreversible actions.

Related Concepts

• Prompt Injection — An attack where untrusted input (a webpage, email, tool output) contains instructions that hijack an LLM's behavior in an agentic context.
• Computer Use — An agent capability where an LLM controls a computer GUI by viewing screenshots and emitting mouse/keyboard actions, automating arbitrary software.