Code and Data
Last updated October 28, 2026
Classgrid AI has its own safe code sandbox. When a job needs real work, such as cleaning a spreadsheet, drawing a chart, doing maths or building a file, the agent writes code, runs it, checks the result and gives you the output. You do not need to know how to code.
The code sandbox
The agent can write and run Python, JavaScript (Node 20) or bash (shell) code. Each run happens in a separate, safe sandbox, and you can watch the steps in the chat.
The sandbox comes ready with common tools, for example:
Data cleaning and analysis
Upload a CSV or Excel file (see Files and documents) and ask in plain words:
- "Clean up messy spreadsheet data: fix the dates, remove duplicates and split the name column."
- "Show total sales by month and the top 5 products."
- "Find customers whose orders dropped by more than 10% since last quarter."
- "Group these survey answers into themes and count each one."
- "Pull the stock price history for this company and show the 50-day average."
The agent loads the data, does the work in code and shows you the result. Ask for the cleaned file back as Excel or CSV.
Charts
There are two kinds of charts:
- Chart cards in the chat — bar, line, pie, doughnut and radar charts you can view right in the answer. See Answer cards.
- Chart images made in the sandbox — useful when you need a picture to put in a report, slide or post.
Try: "Draw a chart from your data", "Make a line chart of monthly revenue and save it as an image", or "Plot this maths function and show the working."
Making files
The agent can create real files and give you a download link:
Files you get from the agent are kept in your Library. Once your WhatsApp is verified, you can also ask "Send the PDF to my WhatsApp", and it arrives as a real file. Or ask the agent to email it (see Email the agent).
A good pattern is: upload data, ask for analysis, then say "Make this into a PDF report with the charts". You get the numbers, the visuals and a file you can share, in one chat.
Coding help
- "Write SQL queries from plain English."
- "Debug an error in your code." (paste the error and the code)
- "Write a Python script that renames these files by date, and test it."
- "Explain what this JavaScript function does, line by line."
The agent can run the code to check it works before giving it to you. For working on a full code project with GitHub, use Project Mode.
Tips
- Describe the result, not the code. "A table of average sales by region" is enough.
- Say the file type you want back. "As Excel", "as a PDF", "as a PowerPoint".
- Check the steps. The chat shows what the agent ran, so you can see how it got the answer.
- Ask it to explain. "Explain what you did in simple words" is useful for learning.
- Keep going. "Now add a column for profit" or "make the chart blue" — it builds on what it already did.