What can Chromeflow actually do?
Concrete workflows AI coding agents (Claude Code, Codex CLI) handle end-to-end with Chromeflow — the parts of your job that used to involve flipping to a browser, doing the click-fill-paste dance, and flipping back to your terminal.
Each page below is a tangible walk-through: which Chromeflow tools the agent reaches for, where the human handoff happens (passwords, 2FA, payment), and what you actually get at the end. Not screenshots and marketing copy — the real tool-call patterns.
Set up a Stripe product + capture the secret key
Agent creates a new Stripe product, sets pricing, configures webhook endpoints, reveals the secret key, and writes it straight to your project's .env. You stay in the loop for the password and 2FA.
Download authenticated Canvas attachments
Canvas attachments live behind student SSO. Chromeflow's privileged-context tools fetch them with your existing session cookies, save to disk, and parse .docx in-extension. No more "log in again to read this PDF".
Capture any API key directly to .env
The pattern that applies to Stripe, Supabase, OpenAI, Anthropic, AWS, SendGrid, Twilio — every SaaS dashboard where the key is revealed once and you have to grab it before navigating away. Agent reveals, captures, writes, confirms.
What ties them all together
All three workflows have the same shape: they live behind a login. Playwright, Browser Use, and Puppeteer launch a fresh-profile browser with no inherited sessions, so each of these tasks would start with re-engineering the login (and silently fail at 2FA). Chromeflow drives your real Chrome where you're already signed in — so the agent's first tool call is the actual work, not the auth dance.
If your agent's job lives behind a login, see how Chromeflow compares to Playwright, Browser Use, and Puppeteer. If you're already convinced, install in two commands from the homepage.