Cuttle
Description
cuttle is a browser for your agent. Websites do not block it. It keeps your logins. When a website asks for a person, you can take control. It works with playwright-cli, agent-browser and browser-use.
Installation
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open the source below and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
README
cuttle
Your agent's stock browser fails on the real web. Sites show captchas, Cloudflare walls and "verify you are human" pages. The browser forgets every login when the process ends. You cannot see what the agent sees. cuttle is a browser for your agent that fixes these three problems:
- Not blocked. The browser is a patched Chromium build that looks like a normal person's browser, with one consistent identity: fingerprint, proxy, IP location, locale and timezone.
- Stays signed in. Sign in once through the viewer. The login persists across agent sessions and across restarts.
- You can step in. A built-in viewer shows the live browser. When a site wants a human, you solve the captcha or 2FA yourself and the agent continues.
It works with playwright-cli, agent-browser and browser-use.
Why not Claude in Chrome or ChatGPT?
They overlap a lot: all three use a real browser, keep your logins, pass bot walls, and read pages fast. The difference is control. As of 2026-08:
| cuttle | Claude in Chrome | ChatGPT | |
|---|---|---|---|
| Banking, tax, government, work accounts | your call | discouraged by policy | gated, auto-pauses |
| Cookies, response bodies, replay from your code | yes | no | no |
| Identities at once | one per container (pool mode for many) | one profile, manual switch | one shared profile |
| Who can drive it | any CDP client | Claude only | ChatGPT only |
| Where it runs | your Docker, server or k8s | your desktop Chrome | OpenAI's cloud |
| Take over when a site pushes back | any browser or phone | at that desk | desktop web only |
The row that matters most is the second one. Reading the full cookie jar, seeing request and response bodies, and calling any origin is what lets an agent work out a site's own API and then replay it from your code. Neither extension exposes that.
Not a reason to choose cuttle: persistent logins, bot walls, page-reading speed, password-manager sign-in. All three do those.
**How it works.** cu
Related Skills
Spec Kit
💫 Toolkit to help you get started with Spec-Driven Development
Testing Webapp Testing
Test local web applications using Playwright for UI verification and debugging
Testing #29
, [#52](https://github.com/affaan-m/everything-claude-code/issues/52), [#103](https://github.com/affaan-m/ever
Testing Fix Issue
by metabase - Addresses GitHub issues by taking issue number as parameter, analyzing context, implementing sol
Testing Pypict Test Design
Design comprehensive test cases using PICT (Pairwise Independent Combinatorial Testing) for optimized test sui
Testing gstack
| 15,000+ | Garry Tan's exact Claude Code setup: 6 opinionated tools that serve as CEO, Eng Manager, Release M
Testing