engineering / skill

dev-browser

A CLI and bundled skill for Chrome automation with short JavaScript scripts and persistent browser pages.

Repository owner SawyerHood

Not rated yet
0 comments

Like 0 Save

Published 11 Oct 2026Listing updated 12 Oct 2026

At a glance

Purpose: A CLI and bundled skill for Chrome automation with short JavaScript scripts and persistent browser pages.
Runs in: Local agent with shell, MCP server
Operating systems: macOS, Linux
Needs: Runtime: Chrome; Needs internet: Yes; Needs accounts or API keys beyond your agent: Not stated
Files and data: Reads files: Yes; Writes files: Yes; Sends data to other services: Not stated
Tool type: Skill
Delivery: From GitHub
Version: Latest on the main branch
Access: Code on GitHub
Compatibility: According to the README, macOS and glibc Linux are supported, and Windows and musl Linux are not yet supported. The tool controls Chrome. The bundled skill installs for Codex, Claude Code or an agents folder. The README shows a Claude Code MCP command and a legacy Claude Code plugin install.
External costs: The repository is MIT-licensed and states no pricing. Your coding agent or model provider may charge for usage.
Permissions: The CLI and skill act through your coding agent with the access you give it. The README says it can start an isolated Chrome profile or connect to a browser you already have open. The README shows global npm installation and skill writes under ~/.codex/skills, ~/.claude/skills and ~/.agents/skills. The README describes downloading the release binary when needed and installing Chrome if the CLI cannot find it. The file list includes a scripts/ directory; the README does not describe its contents. The README says scripts run inside a daemon and can navigate pages, capture snapshots, click, fill fields and read or save files under ~/.dev-browser/v1/tmp. The README describes browser profiles, cookies and login state stored on disk, with configuration under ~/.dev-browser/v1/config.json. The README also shows project or user-level Claude Code settings edits that allow matching dev-browser commands without further prompts and warns that node:vm is not a security sandbox.
Usage terms: MIT License, as stated in the repository. See its LICENSE file for the exact terms.

Code on GitHub

SawyerHood/dev-browser

MIT

TypeScript6,660 stars435 forksLast pushed 5 Sept 2026 UTC

A Claude Skill to give your agent the ability to use a web browser

Topics: claude-code playwright skills

Public GitHub data fetched 13 Oct 2026 UTC · 0 days old

Skill folders in this repository (1)

Folders that contain a SKILL.md file, from GitHub's file list on 13 Oct 2026 UTC.

skills/dev-browser

View on GitHub ↗

What it does

dev-browser is a command-line tool with a bundled skill for Chrome automation. The README says scripts use JavaScript and the browser stays open between calls. The README describes navigating, inspecting pages, acting and checking results across scripts.

The README lists named persistent tabs, accessibility snapshots with element references, the Puppeteer Page API and agent-focused helpers. The README describes launching an isolated Chrome profile or attaching to an existing browser, and an MCP server mode.

The README shows dev-browser install-skill with --codex, --claude and --agents for the corresponding user-level skill directories. The README also suggests asking the agent to run dev-browser --help for its built-in guide.

Who it is for

Developers who want a coding agent to drive Chrome from short JavaScript scripts.

Example tasks

Open a page in a headless browser and print its title.
List the pages open in a Chrome started with dev-browser chrome.
Take an accessibility snapshot of a page and click an element by its ref.
Fill in a form field, then wait for the page to settle.
Connect to a browser you already have open and list its pages.

Getting started

Install the CLI with npm install -g dev-browser. Run dev-browser install only if dev-browser cannot find Chrome. To give an agent the bundled skill, the README shows dev-browser install-skill --codex, dev-browser install-skill --claude and dev-browser install-skill --agents. It says running dev-browser install-skill without flags updates all three locations. For MCP, it shows claude mcp add dev-browser -- dev-browser mcp --headless. The README has the exact commands.

Limitations

According to the README, macOS and glibc Linux are supported. Windows and musl Linux are not yet supported. The tool controls Chrome. The README says to run dev-browser install if dev-browser cannot find Chrome. It says Chrome 136 and newer ignore remote-debugging flags on the default profile, so dev-browser chrome uses a dedicated profile. Snap-packaged Chromium cannot read the default ~/.dev-browser/v1 directory. On Ubuntu, the README says to use dev-browser install or point DEV_BROWSER_CHROME at another Chrome binary. You install with npm. The release binary includes Bun and Puppeteer, so the README says Node is not required after installation. The README says each script runs in a fresh node:vm context that is not a security boundary. The README says launched browsers close after 30 minutes without a script by default; the timeout is configurable, and attached browsers are excluded from idle cleanup. The README warns that concurrent scripts using the same named page can interleave. The README says version 0.2 scripts need helper changes for version 1.0, and page.cua and page.domCua are not part of 1.0. The repository can change at any time; check its README before you install.

Data practices

The README says state lives under ~/.dev-browser/v1 and profiles, cookies and login state remain on disk after a launched browser closes. The README describes credential redaction in logs and status output. The README says the CLI may download its binary on first use and names DEV_BROWSER_DOWNLOAD_BASE for a private mirror and DEV_BROWSER_SKIP_DOWNLOAD=1 for a manually installed binary. The README does not mention telemetry. What else is sent where depends on your coding agent and model provider.