browse.sh icon

browse.sh

browse.sh is an open catalog of browser automation skills and a CLI for AI agents to control websites, inspect sessions, and use local or cloud browser workflows.

browse.sh

What is browse.sh?

browse.sh is an open catalog of browser automation skills and a browser CLI for AI agents. It combines reusable website-specific skills with lower-level browser primitives, debugging tools, and cloud sessions so agents can interact with websites and web apps in a structured way.

The catalog is organized around SKILL.md recipes for specific domains and tasks. The site also shows example commands for adding skills, controlling pages through selectors and accessibility references, inspecting network and console output, and switching between local Chromium and remote Browserbase sessions.

Key Features

  • Open catalog of browser automation skills for specific websites and tasks, with entries such as contract searches, class booking, flight status, reviews, and trail search.
  • browse skills add workflow for installing reusable skill recipes that teach AI agents how to complete website actions.
  • Low-level browser controls including click, type, select, press, hover, scroll, and mouse actions for direct page interaction.
  • Debugging commands for tailing network and console output during a session, which helps inspect requests, responses, warnings, and runtime errors.
  • Support for local Chromium by default, with an option to prefix commands with cloud to use remote sessions and Browserbase APIs.
  • Structured outputs from site-specific skills, such as canonical URLs, status fields, time windows, ratings, or other page data shown in the catalog examples.

How to Use browse.sh

Start by installing the CLI with npm, then use browse skills add to install the skills relevant to your target websites. After that, drive pages with the browser commands for clicking, typing, selecting, and scrolling, or use the catalog entries as ready-made recipes for known sites.

If you need to inspect behavior, tail the network or console for the active session. For remote workflows, use the cloud prefix to create a Browserbase session or call its search and fetch APIs.

Use Cases

  • An AI agent needs a domain-specific recipe to complete a repetitive website workflow, such as booking a class or searching a travel site.
  • A developer wants to automate a browser task while keeping control over page interactions through explicit commands and selectors.
  • A user is debugging a web app and needs to watch network calls and console output while the session runs.
  • A workflow needs to move from local browser automation to a remote session without changing the overall command style.
  • A team wants a reusable skill catalog that can be shared across agents instead of rebuilding prompts for each website.

FAQ

What does browse.sh provide? It provides an open catalog of browser automation skills plus a CLI for running browser actions, debugging sessions, and cloud-based workflows.

Does it work only with local browsers? No. The page says commands work natively with local Chromium, and remote sessions are available by prefixing commands with cloud.

What is a browser skill in this context? A skill is a reusable recipe, described as SKILL.md, that teaches an AI agent how to complete a task on a specific website.

Does browse.sh support debugging? Yes. The site highlights network and console tailing so agents and humans can observe what the page is doing in real time.

Are the catalog entries all interactive tools? Not necessarily. The page shows a mix of API-backed, browser-based, and hybrid entries, so the interaction mode depends on the specific skill.

Alternatives

  • General browser automation frameworks such as Playwright or Puppeteer, which are oriented around scripting browser behavior directly rather than shipping a catalog of reusable skills.
  • Agent/browser orchestration tools that focus on taking natural-language instructions and translating them into web actions, often without a public skill marketplace.
  • Task-specific browser bots or scraping workflows, which may solve one site or one workflow well but do not provide a shared catalog of reusable site recipes.
  • Cloud browser platforms, which emphasize hosted browser infrastructure and session management, while browse.sh combines browser control with a skill catalog and CLI workflow.