Code & Development
Crew can write and execute Python code, analyze data, create visualizations, and even make changes to your GitHub repositories.
Python Sandbox
Crew runs Python code in a secure, isolated sandbox environment. Each execution starts fresh with a clean environment.
Data Analysis
"Analyze this CSV file — find the top 10 customers by revenue, calculate monthly trends, and create a visualization."
"Read the Excel spreadsheet and create a pivot table showing sales by region and product category."
"Calculate the correlation between marketing spend and revenue from this dataset."
Chart Creation
"Create a bar chart showing monthly revenue from this data: [paste or upload data]."
"Generate a line chart comparing our three product lines over the past year."
"Create a pie chart of our expense breakdown."
Data Processing
"Clean this CSV — remove duplicates, fix date formats, handle missing values, and save the cleaned version."
"Merge these two spreadsheets on the 'Customer ID' column and create a combined report."
"Convert this JSON file to a well-formatted CSV."
Pre-installed Libraries
The sandbox comes with these Python libraries ready to use:
pandas — Data manipulation and analysis
numpy — Numerical computing
matplotlib — Chart and visualization creation
openpyxl — Excel file reading/writing
python-docx — Word document handling
Pillow — Image processing
lxml — XML/HTML parsing
How It Works
You ask Crew to perform a data task
Crew writes Python code
The code runs in an isolated sandbox
Results (text output, files, charts) are returned to you
Sandbox Limits
Free
30 seconds
Basic
120 seconds
Plus
300 seconds
Pro
300 seconds
Working with Files
When you upload a file before asking for code execution, the file is automatically available in the sandbox:
[Upload sales_data.xlsx] "Analyze this spreadsheet and create a summary report with charts."
Output files (charts, processed data, generated documents) are saved to My Files.
GitHub Integration
Plan requirement: Pro plan
Repository Management
"List my GitHub repositories."
"Show me the open issues in my-org/my-repo."
"What PRs are waiting for review?"
API Access
Crew has full access to the GitHub API, so you can:
Create and manage issues
Review pull requests
Check CI/CD status
Manage repository settings
And anything else the GitHub API supports
Coding Agent
The Coding Agent can autonomously make changes to your GitHub repositories:
"In my-org/frontend-app, fix the broken login form. The submit button doesn't work on mobile."
"Add input validation to the signup form in my-org/web-app."
"Refactor the database module in my-org/api-server to use connection pooling."
How It Works
You describe the change you want
Crew clones your repository in a secure environment
The agent reads the code, understands the structure, and makes changes
Changes are committed to a new branch
A pull request is created for your review
The agent runs in the background — you'll be notified when the PR is ready.
Best Practices
Be specific: Describe the exact behavior you want, not just "fix the bug"
Reference files: If you know which files need changes, mention them
Always review: The agent creates PRs, never pushes to main — always review before merging
Practical Scenarios
Data Analysis Pipeline
[Upload quarterly_data.xlsx] "Analyze this quarterly data: 1) Calculate year-over-year growth for each product line, 2) Create charts for the top 3 metrics, 3) Generate a PDF report with all findings."
Quick Calculation
"Calculate compound interest on $10,000 at 5% annual rate for 10 years, compounded monthly. Show a table of yearly balances."
File Format Conversion
"Convert this JSON API response to a clean Excel spreadsheet with proper column headers."
Web Scraping + Analysis
"Search the web for the Fortune 500 list, extract the company names and revenues, and create a chart showing the top 20 by revenue."
Automated Reporting
"Set up a weekly task: every Monday at 9 AM, pull the latest data from [API URL], analyze trends, create a chart, and send the report to my email."
Tips
Upload files first: Attach spreadsheets and data files before asking for analysis
Request specific output: "Create a chart" is good; "Create a bar chart with monthly revenue on the Y-axis and months on the X-axis, colored by product line" is better
Iterate: If the first analysis isn't quite right, ask Crew to adjust — it remembers the context
Save results: Ask Crew to save important outputs to My Files or Google Drive
Combine with email: "Analyze the data, create a report, and email it to the team"
Need more depth? For complex multi-step analysis or research that needs polished document output, try Deep Work
Last updated