Code Execution

Code execution lets you run Python scripts that call MonoClaw tools programmatically. This is powerful for collapsing multi-step workflows into a single turn.

When to use code execution

• Batch operations — Process multiple files at once
• Complex logic — Calculations, data transformations
• API orchestration — Chain multiple tool calls with custom logic
• Automation — Scripts that run periodically

Basic usage

In chat, use the code block with python:

```python
# Mona will execute this
import os
files = os.listdir("src/")
print(f"Found {len(files)} files")
for f in files:
    print(f"  - {f}")
```

Accessing tools

Scripts have access to Mona's tools via the tools object:

# Read a file
content = tools.file_read("README.md")

# Run a command
result = tools.terminal("git log --oneline -5")

# Search the web
results = tools.web_search("MonoClaw AI secretary")

Return values

Scripts can return structured data:

files = os.listdir("src/")
{
    "count": len(files),
    "files": files,
    "largest": max(files, key=lambda f: os.path.getsize(f"src/{f}"))
}

Mona receives this data and can use it in her response.

Error handling

Script errors are caught and reported:

try:
    result = tools.terminal(" risky_command ")
except Exception as e:
    print(f"Error: {e}")

File system access

Scripts run with access to:

• The current working directory
• /tmp/ for temporary files
• Read access to mounted volumes (Docker backend)

Security

• Scripts run in the same sandbox as terminal commands
• Dangerous operations require approval (if approval mode is enabled)
• Network access depends on the backend configuration
• File system access is scoped to the working directory

Best practices

• Keep scripts focused — One task per script
• Handle errors — Use try/except for robustness
• Return structured data — Makes it easier for Mona to use results
• Don't hardcode secrets — Use environment variables
• Clean up temp files — Delete /tmp/ files when done

Comparison with skills