Agent Browser — Claude Code skill: headless browser for agents
A skill that hands your agent a Rust-based headless browser: navigate, click, fill forms and snapshot pages through structured commands.
npx skills add am-will/codex-skills --skill agent-browser -a claude-codeWhat Agent Browser does
Agent Browser is a browser automation CLI the agent drives like any other command. Navigation, clicks, typing, extraction and screenshots go through agent-browser open, click, fill and snapshot. The core is written in Rust, with a Node.js fallback.
It works off snapshots and element refs. You take a snapshot, get a list of elements tagged with refs like @e2, then act on them. Refs stay stable for one page load and reset on navigation, so snapshot again after every move. Add --json for machine-readable output, --headed to watch the window while debugging.
For OAuth and logged-in sites the skill either connects to your running Chrome over CDP or keeps a separate session: agent-browser --cdp 9222 open ... reuses your existing profile, while --session persists cookies between runs. Installing needs node and npm. On Linux ARM64 the command may not resolve in PATH — call the binary by its full path inside the bin folder.
What it helps with
- walk through a site and pull structured data without clicking by hand
- fill in a form and submit it from the agent
- snapshot a page and locate the right element by ref
- log in through Google OAuth and keep working in that session
- capture a screenshot or PDF mid-task
Example prompts for the agent
- Open example.com, take a snapshot and list the elements you can act on
- Fill field @e3 with [email protected] and click submit
- Connect to Chrome on port 9222 and open star-swap.com under my session
How to install Agent Browser
1Skills CLI (Vercel)
npx skills add am-will/codex-skills --skill agent-browser -a claude-code2Manually: put the files in place
~/.claude/skills/agent-browser/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/agent-browser/
1Skills CLI (Vercel)
npx skills add am-will/codex-skills --skill agent-browser -a codex2Manually: put the files in place
.agents/skills/agent-browser/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add am-will/codex-skills --skill agent-browser -a cursor2Manually: put the files in place
.cursor/skills/agent-browser/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install am-will/codex-skills agent-browser2Manually: put the files in place
.github/skills/agent-browser/SKILL.md1Skills CLI (Vercel)
npx skills add am-will/codex-skills --skill agent-browser -a gemini-cli2Platform CLI
gemini skills install https://github.com/am-will/codex-skills --path skills/agent-browserManually: put the files in place
.devin/skills/agent-browser/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add am-will/codex-skills --skill agent-browser -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about Agent Browser
How do I install Agent Browser in Claude?
Run npx skills add am-will/codex-skills --skill agent-browser -a claude-code. The skill lands in ~/.claude/skills/agent-browser/SKILL.md.
What does Agent Browser do?
It lets the agent drive a headless browser: open pages, click, type, extract content and take screenshots through structured CLI commands.
How does it handle sites that require login?
Connect to your own Chrome over CDP (agent-browser --cdp 9222 open ...), or use --session so cookies survive between runs after a manual login.
