# Code Execution Tool
The code execution tool enables Grok to write and execute Python code in real-time, dramatically expanding its capabilities beyond text generation.
## Overview
This powerful feature allows Grok to:
- Perform precise calculations
- Complex data analysis
- Statistical computations
- Solve mathematical problems that would be impossible through text alone
- Generate and test code
Code execution happens in a **secure, isolated environment**.
## Key Use Cases
### Mathematical Computations
```python
response = client.chat.completions.create(
model="grok-4",
messages=[
{"role": "user", "content": "Calculate the compound interest on $10,000 at 5% for 10 years"}
],
tools=[{"type": "code_execution"}]
)
```
### Data Analysis
```python
response = client.chat.completions.create(
model="grok-4",
messages=[
{"role": "user", "content": "Analyze this CSV data and find trends: [data]"}
],
tools=[{"type": "code_execution"}]
)
```
### Financial Modeling
Complex financial calculations, portfolio analysis, risk assessment.
### Scientific Computing
Statistical analysis, numerical methods, simulations.
### Code Generation & Testing
Write code and verify it works by executing it.
## Enabling Code Execution
```python
from openai import OpenAI
client = OpenAI(
api_key=os.environ.get("XAI_API_KEY"),
base_url="https://api.x.ai/v1"
)
response = client.chat.completions.create(
model="grok-4",
messages=[
{"role": "user", "content": "Calculate the first 20 Fibonacci numbers"}
],
tools=[
{"type": "code_execution"}
]
)
print(response.choices[0].message.content)
```
## Available Libraries
The sandboxed environment includes common Python libraries:
- **numpy** - Numerical computing
- **pandas** - Data manipulation
- **matplotlib** - Visualization
- **scipy** - Scientific computing
- **sympy** - Symbolic mathematics
- **requests** - HTTP requests (limited)
## Response Format
When code execution is used, the response includes execution results:
```json
{
"choices": [
{
"message": {
"role": "assistant",
"content": "Here are the first 20 Fibonacci numbers: [0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89, 144, 233, 377, 610, 987, 1597, 2584, 4181]",
"tool_calls": [
{
"type": "code_execution",
"code_execution": {
"code": "def fibonacci(n):\n fibs = [0, 1]\n for i in range(2, n):\n fibs.append(fibs[-1] + fibs[-2])\n return fibs[:n]\n\nprint(fibonacci(20))",
"output": "[0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89, 144, 233, 377, 610, 987, 1597, 2584, 4181]"
}
}
]
}
}
]
}
```
## Combining with Other Tools
Code execution works well with other tools:
```python
response = client.chat.completions.create(
model="grok-4",
messages=[
{"role": "user", "content": "Search for Bitcoin price history and plot it"}
],
tools=[
{"type": "web_search"},
{"type": "code_execution"}
]
)
```
## Visualization
The model can generate visualizations using matplotlib:
```python
response = client.chat.completions.create(
model="grok-4",
messages=[
{"role": "user", "content": "Create a bar chart showing the top 5 programming languages by popularity"}
],
tools=[{"type": "code_execution"}]
)
```
## Security
- Code runs in an isolated sandbox
- No access to the host filesystem
- Limited network access
- Execution time limits
- Memory limits
## Best Practices
1. **Be specific**: Clearly describe what calculation or analysis you need
2. **Provide data**: Include data inline or reference attached files
3. **Request visualization**: Ask for charts when visual output is helpful
4. **Verify results**: Review the generated code and output
5. **Iterative refinement**: Ask follow-up questions to refine analysis
## Limitations
- Execution time limits apply
- Some libraries may not be available
- Network requests are restricted
- Cannot persist data between executions
- File system access is sandboxed
