Overview

When ChatGPT executes Python code using the Run feature, it operates inside a temporary sandbox environment.
The primary working directory available to scripts inside that environment is:

/mnt/data

This directory functions as the temporary filesystem workspace for:

  • uploaded files
  • generated files
  • intermediate data during processing
  • scripts written during a session

Understanding how /mnt/data works is important for workflows involving file processing, data transformation, and temporary automation tasks.

What /mnt/data Is

/mnt/data is a temporary storage directory inside the sandbox container used by ChatGPT’s Python runtime.

Characteristics:

  • Writable
  • Isolated from the internet
  • Accessible to Python scripts
  • Not connected to your local computer
  • Automatically cleaned up after the session ends

Typical contents may include:

/mnt/data/input.csv
/mnt/data/report.xlsx
/mnt/data/image.png

Files appear here when you:

  • upload files into the chat
  • run scripts that generate files
  • download artifacts created during execution

Scope of the Directory

The scope of /mnt/data is limited to the current execution environment.

The environment is:

  • sandboxed
  • ephemeral
  • isolated per session

The directory cannot access:

  • your local filesystem (for example C:\Users\)
  • OneDrive
  • network shares
  • external APIs that require filesystem access

It exists only within the temporary Python runtime container.

Lifetime of Files

Files stored in /mnt/data persist only during the active chat session.

Typical behavior:

Event Result ———————— ———————— Upload file Appears in /mnt/data Generate output file Saved in /mnt/data Run additional scripts Files remain available Long idle period Files may disappear Chat session ends Files are deleted

Typical persistence window:

  • 30–60 minutes of active work
  • Not guaranteed across container resets

Always download important files immediately after generation.

Typical Workflow

A common processing workflow looks like this:

  1. Upload files into the chat
  2. Files appear in /mnt/data
  3. Run Python script
  4. Script reads input files
  5. Script writes output files
  6. Download generated results

Example:

Upload → /mnt/data/statements.pdf
Run script → /mnt/data/summary.xlsx
Download → summary.xlsx

Listing the Directory

You can inspect the directory contents using Python.

Example script:

import os

for root, dirs, files in os.walk("/mnt/data"):
    for f in files:
        print(os.path.join(root, f))

Example output:

/mnt/data/file1.csv
/mnt/data/file2.pdf
/mnt/data/output.xlsx

Detailed Directory Listing

A more advanced listing script:

import os
from datetime import datetime

BASE = "/mnt/data"

for root, dirs, files in os.walk(BASE):
    for name in files:
        path = os.path.join(root, name)
        stat = os.stat(path)

        size_kb = stat.st_size / 1024
        modified = datetime.fromtimestamp(stat.st_mtime)

        print(path, size_kb, modified)

This shows:

  • full path
  • file size
  • last modification timestamp

Common Operations

Typical file operations supported inside /mnt/data.

Read a file

with open("/mnt/data/input.txt") as f:
    data = f.read()

Write a file

with open("/mnt/data/output.txt","w") as f:
    f.write("Hello world")

Save a pandas dataframe

df.to_excel("/mnt/data/report.xlsx", index=False)

Delete a file

import os
os.remove("/mnt/data/temp.txt")

Useful Utilities

Show only PDFs

if name.lower().endswith(".pdf"):

Calculate total disk usage

total = sum(
    os.path.getsize(os.path.join(root,f))
    for root,_,files in os.walk("/mnt/data")
    for f in files
)
print(total/1024/1024,"MB")

Show directory tree

for root, dirs, files in os.walk("/mnt/data"):
    print(root)

Best Practices

Recommended usage guidelines.

  1. Treat /mnt/data as scratch storage
  2. Download results immediately
  3. Avoid relying on persistence
  4. Use small to moderate file sizes
  5. Re-upload files if the environment resets

When to Use /mnt/data

Good use cases:

  • CSV → XLSX conversions
  • PDF data extraction
  • quick data cleanup
  • visualization
  • temporary automation tasks

Not recommended for:

  • long-term storage
  • large datasets
  • production pipelines

Relationship to Local Development

For repeated workflows, scripts should live in a permanent environment such as:

  • a local Python installation
  • a Git repository
  • an automated data pipeline

The /mnt/data directory is best viewed as:

temporary sandbox workspace

rather than a permanent filesystem.

Summary

/mnt/data is the temporary filesystem used by ChatGPT’s Python runtime.

Key points:

  • temporary sandbox directory
  • holds uploaded and generated files
  • accessible only within the current session
  • cleared when the environment resets
  • ideal for quick data processing workflows

Understanding this directory allows you to efficiently use ChatGPT as a temporary data-processing environment and rapid scripting tool.