Code Execution (Claude)
The Code Execution (Claude) tool (anthropic_code_execution) enables Claude to run code in a secure sandbox. This makes calculations, data analysis and automation possible.
Only in assistants, lessons and workflows
Code execution is always off in standard chats, even if the environment or your personal settings say otherwise. Switch the tool on in the tool settings of an assistant, lesson or workflow that needs calculations or file editing. Code execution can take longer and add costs.
How it works
Claude can write code and run it immediately for calculations, analyses and file processing. The results are integrated into the answer.
Usage examples
- "Calculate the monthly depreciation for this asset"
- "Analyse this dataset and identify trends"
- "Create a chart of our quarterly figures"
- "Process this CSV file and generate a report"
Settings
| Setting | Value |
|---|---|
| Tool key | anthropic_code_execution |
| Default status | Auto |
| Available for | Suitable Claude models |
Capabilities
Calculations
- Financial calculations (ROI, NPV, IRR)
- Statistical analyses
- Mathematical models
- Scenario analyses
Data analysis
- CSV and Excel processing
- Trend analysis
- Aggregations and summaries
- Data transformations
Visualisations
- Graphs and charts
- Dashboards
- Reports with visuals
Business applications
- Finance: Financial models, budgeting, forecasting
- Sales: Pipeline analysis, forecast models
- Operations: KPI calculations, efficiency analyses
- HR: Headcount planning, staff turnover figures
Security
- Isolated execution environment
- No access to external systems
- Temporary file storage
- Enterprise-grade security
Related tools
- Code Interpreter (OpenAI) - Similar for OpenAI
- Wolfram Alpha - For specific calculations