How to Use Code Interpreter on Magicdoor

Code interpreter lets you run Python code directly inside your Magicdoor chat. Upload a spreadsheet, ask a question about it, and the model writes and executes real Python to give you an answer — complete with charts, processed files, and data summaries. No setup, no local environment, no coding knowledge required.

What Code Interpreter Does

When you upload a data file or ask a question that benefits from computation, Magicdoor runs Python in a secure sandbox behind the scenes. The model can:

• Analyze data files — CSV, Excel, JSON, and more
• Create visualizations — bar charts, line graphs, scatter plots, histograms
• Process and transform data — filter, sort, aggregate, merge datasets
• Perform calculations — statistics, financial modeling, unit conversions
• Generate output files — cleaned CSVs, processed spreadsheets, chart images

You see the results directly in the chat. Generated files (charts, processed data) appear as downloadable items in the conversation.

Which Models Support It

Code interpreter is available on two OpenAI models:

If you upload a file that needs code interpreter (like a CSV or Excel file) while using a different model, Magicdoor automatically switches to a compatible model for you.

Step-by-Step: Analyze a Spreadsheet

1. Start a chat and upload your file

Open the chat interface and click the attachment button (or drag and drop) to upload your file. Supported file types include:

• Spreadsheets: CSV, XLSX, XLS
• Data files: JSON, XML, TXT
• Documents: DOC, DOCX, PPTX
• Code files: PY, JS, TS, HTML, CSS, and more
• Archives: ZIP, TAR

2. Ask your question in plain language

You do not need to write code. Just describe what you want:

• "Summarize this dataset — what are the key trends?"
• "Create a bar chart of sales by region"
• "Find all rows where revenue is above $10,000 and export them as a new CSV"
• "What's the average, median, and standard deviation of the 'price' column?"

3. Review the results

The model will:

1. Explore your data (column types, row count, sample values)
2. Write and run Python code to answer your question
3. Show you the results — numbers, tables, or charts — directly in chat
4. Make generated files (charts, processed data) available for download

4. Ask follow-up questions

The sandbox persists throughout the conversation, so your data stays loaded. Ask follow-ups without re-uploading:

• "Now break that chart down by quarter"
• "Remove outliers and recalculate"
• "Export just the top 10 rows as a CSV"

Practical Use Cases

Sales and business data

Upload a sales spreadsheet and ask: "Show me monthly revenue trends for 2025 with a line chart. Highlight any months where revenue dropped more than 10% from the previous month."

Survey analysis

Upload survey responses and ask: "What are the most common answers to question 3? Show a pie chart of the distribution and calculate the margin of error."

Financial modeling

Ask the model to calculate compound interest, loan amortization, or investment returns. Upload your own numbers or describe the scenario.

Cleaning messy data

Upload a CSV with inconsistencies and ask: "Standardize the date formats, remove duplicate rows, fill in missing values with the column median, and export a clean version."

Code review and testing

Upload a Python file and ask: "Run this script and show me the output" or "Find potential bugs in this code and test the edge cases."

Tips for Better Results

Be specific about what you want. "Analyze this data" is vague. "Show me the correlation between columns A and B as a scatter plot with a trend line" gets you exactly what you need.

Start with exploration. If you are not sure what is in your data, start with "Describe this dataset" or "Show me the first few rows and basic statistics." Then ask targeted questions.

Use GPT-5.4 Mini for simple tasks. If you just need a quick calculation or a basic chart, Mini is 3-6x cheaper than the full GPT-5.4 and handles straightforward analysis well.

Iterate in the same conversation. Each follow-up question builds on the previous context. The model remembers your data, any transformations it made, and the charts it created.

Upload multiple files. You can upload several files in the same conversation. Ask the model to merge, compare, or cross-reference them.

What Code Interpreter Cannot Do

• No internet access — the Python sandbox is isolated and cannot fetch external data or APIs
• No GPU or heavy ML training — it runs standard Python with common data libraries (pandas, matplotlib, numpy, etc.), not heavy ML workloads
• File size limits — very large files (over ~100 MB) may be slow or fail to process
• Session timeout — the sandbox expires after a period of inactivity; for long analysis sessions, download intermediate results

FAQs

Do I need to know Python to use code interpreter?

No. You ask questions in plain English and the model writes all the code. You never need to see or edit the code unless you want to.

How much does code interpreter cost?

You pay the normal token cost for GPT-5.4 or GPT-5.4 Mini — there is no extra charge on Magicdoor for code execution itself. A typical data analysis conversation costs a few cents. For detailed pricing, see our model cost guide.

Can I use code interpreter with Claude or Gemini?

Currently, code interpreter is only available with GPT-5.4 and GPT-5.4 Mini. If you upload a file that needs code interpreter while using a different model, Magicdoor will automatically switch to a compatible model.

What Python libraries are available?

The sandbox includes the most common data science libraries: pandas, numpy, matplotlib, seaborn, scipy, and more. It covers the vast majority of data analysis and visualization tasks.

Can I download the files it generates?

Yes. Charts, processed spreadsheets, and other output files appear in the chat and can be downloaded directly.

Is my data secure?

Your uploaded files are processed in an isolated sandbox that expires after your session. Magicdoor does not use your data for model training, and your content is encrypted in transit and at rest.

Get Started

Upload a CSV or Excel file to any chat on Magicdoor and ask a question about it. That is all it takes. If you are new to Magicdoor, create a free account to try it — no credit card required.

For more ways to use AI effectively, check out our guides on developer workflows, prompt engineering, and multi-model workflows.