agent-browser
FreeBrowser automation CLI designed for AI agents, featuring compact text output, ref-based selection, and native Rust implementation.
Informations sur l’outil
Qu'est-ce que agent-browser ?
agent-browser is a command-line browser automation tool optimized for AI agents. It delivers compact accessibility-tree snapshots with deterministic refs to minimize token usage and enable reliable web interactions via shell commands.
Fonctionnalités principales
Agent-first compact output
Produces compact text output that uses fewer tokens than JSON, designed specifically for AI context efficiency.
Ref-based element selection
Snapshot command returns an accessibility tree with unique refs (such as @e1, @e2) for deterministic element selection and interaction without DOM re-query.
Complete command set
Offers 50+ commands covering navigation, forms, screenshots, network, storage, files, tabs, frames, and debugging.
Built-in observability tools
Includes video recording, streaming, debugging, profiler, and diffing tools.
Modern application support
Provides first-class support and documentation for network control, React & Web Vitals, init scripts, and Next.js + Vercel workflows.
Stateful sessions
Supports sessions, profiles, auth state, cookies, storage, proxy, and security controls for long-running agents.
Cross-platform native binaries
Delivers native Rust binaries for macOS (ARM64, x64), Linux (ARM64, x64), and Windows (x64).
Client-daemon architecture
Uses a Rust CLI client that communicates with a pure Rust native daemon managing Chrome via the Chrome DevTools Protocol; daemon starts automatically and persists between commands.
Cas d'utilisation
- #1
AI agent web automation
Enable AI agents that can run shell commands to open pages, take snapshots, click refs, fill forms, capture screenshots, and close sessions for autonomous browser workflows.
- #2
Token-efficient page understanding
Obtain compact accessibility-tree snapshots that use roughly 200-400 tokens instead of thousands for full DOM, allowing efficient LLM parsing and interaction planning.
- #3
Persistent multi-step agent sessions
Maintain auth state, cookies, profiles, storage, and proxy settings across commands so long-running agents can continue authenticated workflows.
- #4
Debugging and observability of browser tasks
Record video, stream sessions, profile performance, and diff page states while automating modern web applications including React and Next.js sites.
Plans tarifaires
Installable at no cost via npm (global or npx), Homebrew, or by downloading native binaries; no paid plans, tiers, or pricing details are mentioned on the official site.
Alternatives
Foire aux questions
What is agent-browser?
How does agent-browser work with AI agents?
Is agent-browser free to use?
Which platforms does agent-browser support?
What tools and environments does agent-browser integrate with?
agent-browser Contacts et réseaux sociaux
- Site Internet :
- GitHub
Avis
Partagez votre expérience avec cet outil...