Deep Dive Into ChatGPT Work: Persistent Filesystem, Web Browser, and Cloud Deployments
ChatGPT Work offers advanced features missing from the standard chat interface. Key highlights include an internet-connected Python sandbox, a persistent cross-session filesystem, and headless Chrome browser control.
Impact: Medium
Why it matters
You can build and deploy stateful web apps directly to Cloudflare, or write scripts that interact with the web via headless browser sessions right from your phone.
TL;DR
- 01ChatGPT Work Cloud features full, unrestricted internet access for Python script execution.
- 02Filesystem workspace /workspace/scratch persists and syncs file states across parallel browser/mobile sessions.
- 03Built-in headless Chrome via Playwright enables complex interaction, DOM scraping, and 2FA session takeovers.
Key facts
- Cost Tier
- $20/month Plus, Team, Enterprise
- Internet Access in Sandbox
- Unrestricted
- Persistent Folder
- /workspace
Persistent Workspace Shared Folders
Unlike standard ChatGPT sessions, which get isolated filesystems that disappear after closing the tab, ChatGPT Work provides a unified workspace. Every session mounts a persistent /workspace directory containing directories like: /workspace/scratch/e00a0a017944 Edits made to files in one workspace are immediately visible to other concurrent Work sessions, creating a shared scratchpad for complex multi-agent tasks.
Sandbox with Web Access and Headless Chrome
In standard chats, the Python execution container is heavily sandboxed behind proxies that block outbound internet access. In ChatGPT Work, the container can access the public web freely.
The browser tool leverages Playwright to spin up a headless Chrome instance. You can instruct the agent to run JavaScript against any DOM. For example, to scrape headers, ChatGPT Work executes:
await tab.playwright.evaluate(() => {
return Array.from(document.querySelectorAll("h1,h2,h3,h4,h5,h6"), heading => ({
level: heading.tagName.toLowerCase(),
text: heading.innerText.trim().replace(/\s+/g, " ")
}));
});Cloudflare Worker Deployments
ChatGPT Work features built-in capabilities to generate, test, and deploy functional websites on Cloudflare Workers. It can write HTML, JavaScript, and manage stateful server-side storage using:
- Cloudflare D1 (SQL database)
- Cloudflare R2 (Object storage)
These sites are deployed directly to a chatgpt.site subdomain, defaulting to private but sharable with specific team members.
Try it in 2 minutes
await tab.playwright.evaluate(() => {
return Array.from(document.querySelectorAll("h1,h2,h3,h4,h5,h6"), heading => ({
level: heading.tagName.toLowerCase(),
text: heading.innerText.trim().replace(/\s+/g, " ")
}));
});javascript
✓ When to use
- When writing scripts that need to install dynamic npm/pip packages and interact with external APIs.
- When scraping websites that require running custom JavaScript and browser interaction.
- When building and deploying simple stateful prototype websites using Cloudflare Workers.
✕ When NOT to use
- Not for highly sensitive, air-gapped data operations that cannot risk exposure to external third-party environments.
What to do today
- Log into ChatGPT Work and verify access to the persistent volume at /workspace.
- Test running external python requests or cloning a public repository in the execution sandbox.
- Experiment with Playwright scripts inside the Work interface to fetch and parse dynamic site layouts.
Sources