Agent Skills

steel-browser

🔥 The easiest way to run AI Agents from your terminal. Scale your agents with Steel.

インストール
CLI
npx skills add https://github.com/steel-dev/cli --skill steel-browser

CLI を使用してこのスキルをインストールし、ワークスペースで SKILL.md ワークフローの使用を開始します。

最終更新日: 4/22/2026

@steel-dev/cli

Steel CLI for browser automation, API tools, and agent workflows.

This package now integrates agent-browser directly into steel browser, so you can keep familiar browser commands while using Steel-native session lifecycle, auth, and endpoint handling.

Jump To

Install

curl -sSf https://setup.steel.dev | sh

Or download directly from GitHub Releases.

Upgrading from npm

If you previously installed via npm i -g @steel-dev/cli:

npm update -g @steel-dev/cli   # auto-installs native binary
# Then optionally:
npm uninstall -g @steel-dev/cli
export PATH="$HOME/.steel/bin:$PATH"

Quick Start

Cloud mode (default)

steel login
steel browser start --session my-job
steel browser open https://example.com --session my-job
steel browser snapshot -i --session my-job
steel browser stop

Agent-Browser Integration

steel browser is directly backed by the vendored agent-browser runtime.

Steel-owned lifecycle commands:

  • steel browser start
  • steel browser stop
  • steel browser sessions
  • steel browser live

All other steel browser <command> calls are inherited from upstream runtime behavior and routed through Steel.

Migration from upstream agent-browser is typically command-prefix only:

  • before: agent-browser <command> ...
  • after: steel browser <command> ...

Read more:

Command Overview

Group Commands
Quickstart forge, run
Browser lifecycle browser start, browser stop, browser sessions, browser live
Browser passthrough steel browser <inherited-command>
Browser profiles profile import, profile sync, profile list, profile delete
API tools scrape, screenshot, pdf
Local runtime dev install, dev start, dev stop
Account and utility login, logout, config, settings, cache, docs, support, star, update

For full flags and argument schemas, use the generated reference:

Endpoint Resolution

For browser lifecycle, browser passthrough bootstrap, and top-level API tools (scrape, screenshot, pdf), endpoint selection is deterministic.

Self-hosted precedence (highest to lowest):

  1. --api-url <url>
  2. STEEL_BROWSER_API_URL
  3. STEEL_LOCAL_API_URL
  4. browser.apiUrl in ~/.config/steel/config.json
  5. http://localhost:3000/v1

Cloud precedence:

  1. STEEL_API_URL
  2. https://api.steel.dev/v1

Attach-flag override rule:

  • If --cdp or --auto-connect is provided, Steel bootstrap injection is skipped and passthrough args are forwarded unchanged.

Output and Runtime Notes

  • steel scrape defaults to markdown-first output for token efficiency; use --raw for full JSON payload.
  • steel browser start and steel browser sessions emit display-safe connect URLs with sensitive query values redacted.
  • Browser command paths bypass auto-update checks for lower interactive latency.

Documentation Map

Primary docs:

Reference docs:

Agent skill: