#99th of 20 in Agent frameworks and SDKs
agent-browser
Native Rust CLI that lets AI agents drive Chrome via accessibility snapshots
- Stars
- 44k
- License
- Apache-2.0
- Last commit
- Oct 2026
- Last release
- Oct 2026
- Language
- Rust
24,423 on 25 Mar 202643,985 on 14 Oct 2026
From GitHub's daily star counts.
Checked on GitHub.
Vercel is hiring: 90 open roles
Overview
agent-browser is a Rust command-line tool that controls Chrome for AI agents. It exposes an accessibility-tree snapshot with element refs (@e1, @e2) that commands like click, fill and get text can target, alongside CSS and semantic locators. It also covers screenshots, network interception, HAR recording, cookies, batch execution, an optional MCP profile and experimental WebMCP support.
Who it is for: Developers giving AI agents scriptable browser control from the shell
Strengths
- Snapshot refs let agents target elements without writing selectors
- Native Rust binary; daemon needs no Playwright or Node.js
- Batch mode runs multiple commands in one process invocation
- Detects existing Chrome, Brave, Playwright and Puppeteer installs
Weaknesses
- Requires downloading Chrome via a separate install step
- WebMCP support is marked experimental
- Linux needs extra system libraries via install --with-deps
- Refs go stale after page changes and need a fresh snapshot
What it needs
- no GPU
- Docker + Compose
- Needs Chrome (Chrome for Testing), Node.js 24+ and pnpm 11+ (source build only), Rust (source build only)
Also in Agent frameworks and SDKs
See all 20| Rank | Project | Score |
|---|---|---|
| 1 | ppt-masterAgent skill that turns documents into natively editable PowerPoint decks | 90 out of 100 |
| 2 | langchainPython framework for building LLM agents and apps from interoperable components | 88 out of 100 |
| 3 | browser-usePython library that lets an LLM agent control a web browser | 88 out of 100 |
| #4 | tradingagentsLangGraph multi-agent framework simulating a trading firm to analyze tickers | 86 out of 100 |
| #5 | crewaiPython framework for multi-agent workflows with role-based Crews and event-driven Flows | 80 out of 100 |