nicobailon/surf-cli

The CLI for AI agents to control Chrome. Zero config, agent-agnostic, battle-tested.

What it solves

Surf provides a simple, agent-agnostic CLI for AI agents to control a Chrome browser. It eliminates the need for complex setups, proprietary API keys, or specific AI provider lock-in, allowing agents to interact with the web as a human would—using existing browser logins and sessions.

How it works

Surf operates as a CLI tool that communicates with a Chrome extension via a Unix socket. It uses a native messaging host to bridge the CLI and the browser. This architecture allows it to be used by any tool that can execute shell commands (such as Claude Code, GPT, or custom shell scripts).

Key technical features include:

  • Accessibility Tree Integration: Uses stable element references (e1, e2, etc.) from the accessibility tree, making interactions more resilient to DOM changes than traditional CSS selectors.
  • Session Management: Implements a session system to allow multiple independent agents to operate concurrent browser tabs without interfering with each other.
  • Remote Access: Supports remote browser control over Tailscale, using mutual Ed25519 challenge-response for secure authentication.
  • Smart Defaults: Automatically resizes screenshots to save tokens and captures screenshots after actions to reduce round-trips for the agent.

Who it’s for

Developers building AI agents that need to browse the web, perform research, or interact with web applications without the overhead of managing headless browser clusters or complex API integrations.

Highlights

  • Agent-Agnostic: Works with any LLM or agent framework that can run CLI commands.
  • Zero Config: Quick installation via npm and a Chrome extension.
  • Semantic Locators: Find and interact with elements by ARIA role, text, or label.
  • Harnesses Browser Cookies: Query AI services like ChatGPT, Gemini, and Perplexity using existing browser logins instead of API keys.
  • Iframe Support: Ability to switch between frames to interact with nested content.
  • Network Capture: Automatically logs network requests for filtering and replaying API calls.

Related

  • Project
    vercel-labs/agent-browser

    A native Rust CLI that lets AI agents (or scripts) control Chrome via simple commands, using accessibility‑based element refs (`@eN`). It supports opening pages, taking snapshots, clicking/filling, extracting text, screenshots, network mocking, tab management, batch execution, and an experimental WebMCP tool system.

  • Project
    jackwener/OpenCLI

    A tool that converts websites and browser sessions into a CLI, enabling humans and AI agents to automate web interactions using logged-in Chrome profiles.

  • Project
    sleepinginsummer/agent-browser-cli

    A CLI tool that connects AI Agents to a real Chrome browser session, allowing them to reuse login states and perform precise page control and perception.

  • Project
    keon/browser-control

    A fast Rust CLI that lets AI coding agents control a real browser via the Chrome DevTools Protocol using shell-native commands and stable element references.

  • Project
    yeahhe365/Gemini-Nexus

    A Chrome extension that adds a native AI layer to the browser, integrating multiple LLM providers and enabling agentic browser control and external MCP tool support.