sandbox:/ mnt/ data Link Not Working in ChatGPT? How to Get Your File
ChatGPT gave you a sandbox:/
Short answer
sandbox:/mnt/data is a folder inside the temporary Python environment ChatGPT runs on OpenAI's servers, not on your computer. The link works only while that environment is alive. When it has expired you see File not found or a session-expired error, and the file is gone. Ask ChatGPT to recreate the file and download it at once, or ask for the code as text and save it yourself.
Got the files? Put them online in 30 seconds
Drop the ZIP or the HTML from ChatGPT into a new VibeDeploy site and get a live HTTPS address, with your own domain one step away. Works with every ChatGPT plan. 14 days free, no card.
ChatGPT built what you asked for and ended with something like "Download your site here: sandbox:/". You click it and nothing happens. Or you get File not found, or The code interpreter tool session has expired. Pasting the link into your browser does nothing either. The file is not lost because you did something wrong; the link points somewhere you cannot reach. Here is what the path means, why the link dies, and how to get your file, and your site, out.
What is sandbox:/mnt/data?
When ChatGPT writes and runs code for you, for example to build a ZIP, convert a file or generate a website, it does that in a temporary Python environment on OpenAI's servers: a small Linux machine with its own filesystem. / is the working folder of that machine. Files you upload for analysis land there, and so do the files ChatGPT creates.
sandbox: is how ChatGPT refers to that machine in its answer. A link such as sandbox:/ means "the file website.zip in the working folder of my code environment". The ChatGPT interface turns that link into a download while the file still exists. It is not a web address and not a folder on your computer, which is why your browser does not know what to do with it and why searching your disk for mnt finds nothing.
You will see the path in a few forms. They all point to the same place:
sandbox:/, for examplemnt/ data/ <file> sandbox:/ormnt/ data/ index.html sandbox:/: the download link ChatGPT writes.mnt/ data/ website.zip /without the prefix: the same file as it appears in the code ChatGPT runs and in its error messages.mnt/ data/ <file> sandbox:/: the folder itself.mnt/ data
Why the link stops working
The environment expired. The code environment is temporary. After a while without use it is reset, and everything in / disappears with it. OpenAI does not publish how long ChatGPT keeps it. For the code interpreter in its API, OpenAI documents that a container expires after 20 minutes without use and that its data is then discarded and not recoverable, and users of ChatGPT report the same pattern: the link works right after the answer and fails after a break. Clicking an old link then gives "The code interpreter tool session has expired." or "File not found".
The file was never created. Sometimes ChatGPT writes a download link in its answer although the code that should create the file did not run in that turn, or failed. The link looks normal and leads nowhere.
The app does not render the link. Some users report links that work in a desktop browser but not in the mobile app, and others the reverse. If the link shows up as plain text, try the same chat on the other one.
How to get the file back
1. Ask ChatGPT to recreate it, then download at once. In the same chat, send something like:
Run the code again to recreate website.zip in /mnt/data. List the files in /mnt/data so I can see it exists, then give me a new download link.The file listing tells you whether the file really exists this time. Download it as soon as the link appears; do not leave it for later. If the environment was reset, ChatGPT may also have lost the files you uploaded earlier. Upload them again when it asks.
2. Ask for the code as text. This always works, because it does not depend on the environment at all. Ask ChatGPT to print every file of the site in its own code block, with the file name above each block. Save each block in one folder under that name: index.html, styles.css, script.js. Open index.html in your browser to check it. For a one-page site, ask for everything in a single index.html and you only have one file to save.
3. Keep the next one safe. Download files the moment ChatGPT makes them, and for a site with several files ask for one ZIP rather than a link per file.
Got the files? Put them online in 30 seconds
Drop the ZIP or the HTML from ChatGPT into a new VibeDeploy site and get a live HTTPS address, with your own domain one step away. Works with every ChatGPT plan. 14 days free, no card.
From /mnt/data to a live website
Once the files are on your computer, they can go online. Two routes:
ChatGPT Sites. On Plus, Pro and workspace plans, ChatGPT can publish a site itself on OpenAI's hosting, with a custom domain where that is available. OpenAI's help center says Sites is not available on Free or Go.
Your own host, from any ChatGPT plan. On VibeDeploy:
- Start the 14-day trial (no card) and create a site.
- Drop in the ZIP, the folder or the single HTML file. A ZIP that wraps everything in one top-level folder is fine. About 30 seconds later the site is live at
https://.name.vibedeploy.be - Add your own domain in the dashboard and create the DNS records it lists. The certificate is issued automatically.
You can also skip the download: copy the deploy guide from your VibeDeploy dashboard, where it is called the Magic File, paste it into the chat and say "deploy this". Depending on what the chat can do, ChatGPT sends the files itself or gives you a deploy page to open in your browser. Keep the guide out of public files; it holds your deploy key. More on the ChatGPT routes, plan by plan, in deploy a ChatGPT-built website.
One check before you upload: if ChatGPT made a React, Vite or Next.js project, the ZIP holds source code with a package.json, not a finished site. Ask ChatGPT for a plain HTML, CSS and JavaScript version, or build the project yourself (npm run build) and upload the build folder, usually dist/.
Using Claude instead? Its sandbox has the same kind of path: see /home/workdir/artifacts/index.html.
Sources
- OpenAI API documentation, Code Interpreter: "A container expires if it is not used for 20 minutes", after which its data is discarded and not recoverable. This covers the API; OpenAI publishes no timeout for ChatGPT itself. Checked 26 October 2026.
- OpenAI Developer Community, Issue with PDF download link in ChatGPT, "sandbox:/mnt/data" not accessible (opened 3 April 2025): links that do not download, "Not found", links working in the browser but not the app, and regenerating the file for a fresh link as the fix that worked. Checked 26 October 2026.
- OpenAI Developer Community, Download link not working for generated files in ChatGPT code interpreter (opened 5 April 2025): "file not found" on generated files. Checked 26 October 2026.
- OpenAI Developer Community, Issue downloading a PPTX file created with Python in ChatGPT (opened 21 May 2024): "The code interpreter tool session has expired." Checked 26 October 2026.
- OpenAI Help Center, Creating and using ChatGPT Sites: Sites on Plus, Pro and workspace plans, not on Free or Go. Checked 13 October 2026 for the ChatGPT deploy guide on this site.
Ship your AI-built site in minutes
VibeDeploy hosts your AI-built websites in the EU with custom domains, automatic SSL, and a 14-day free trial that gets you online today.
Deploy guides for this topic
Questions people ask
What is sandbox:/mnt/data?
It is how ChatGPT points at a file inside its own code environment: a temporary Linux machine on OpenAI's servers where it runs Python. /mnt/data is the working folder there. In the chat, a sandbox:/mnt/data link turns into a download link while the file still exists. It is not a folder on your computer and not a web address you can open in a browser.
Why does my sandbox:/mnt/data link say File not found?
Because the file no longer exists in that environment. The environment is temporary: after a period without use it is reset and everything in /mnt/data is deleted. OpenAI does not publish ChatGPT's timeout; for the code interpreter in its API it documents that a container expires after 20 minutes without use and its data is discarded and not recoverable. Sometimes the file was never created in the first place.
What does "The code interpreter tool session has expired" mean?
The environment that held the file is gone, so the download cannot start. Nothing on your side fixes that link. Ask ChatGPT in the same chat to run the code again, recreate the file, confirm it exists and give you a new link, then download it straight away.
Can I open sandbox:/mnt/data on my computer?
No. The path exists only inside ChatGPT's environment, so searching your disk or pasting the link into the address bar finds nothing. Download the file through a working link in the chat, or ask ChatGPT to print the contents as code and save them yourself.
How do I put the website ChatGPT made online?
Download the ZIP or HTML file, or save the code as index.html, and upload it to a web host. On VibeDeploy you drop it into a new site and get a live HTTPS address in about 30 seconds, then add your own domain. On Plus, Pro and workspace plans, ChatGPT Sites can also publish it on OpenAI's hosting.
Related reading
How to Deploy AI-Built Websites in 2026: The Complete Guide
Every way to get an AI-built site live in 2026: MCP deploys, a deploy guide for any chat AI, upload or GitHub, containers, and a 25-tool export matrix.
Claude Artifact Won't Publish or Share? Causes and Fixes
Why a Claude artifact won't publish or share: Claude accounts, admin settings, the 16 MiB limit, the renamed button. How to unpublish, and the fix.
/ home/ workdir/ artifacts/ index.html: How to Get Your File Out
Claude saved your site to /
"Uploads the file below from this task to claude.ai": Claude's Publish Artifact Prompt Explained
What Claude's Publish artifact prompt means: "Uploads the file below from this task to claude.ai", who can see the page, the three variants, and how to undo it.