Cappy Code is a PHI-safe agentic code runner (similar to Claude Code) that executes tasks in a controlled, auditable manner. It handles local file operations, shell commands, and more, but maintains guardrails for privacy and compliance.
-
Clone or Copy the Repo
git clone git@github.com:susom/cappy_code.git cd cappy_code -
Install Dependencies
pip install -e .This installs Cappy Code in editable mode. It will also install any required Python libraries from requirements.txt.
-
Set up SecureChatAI Credentials
- Copy
.env.exampleto.env. - Open
.envand set the required environment variables (likeREDCAP_API_URL,REDCAP_API_TOKEN), ensuring you keep these secrets out of source control.
- Copy
- IMPORTANT: You must also have a REDCap API token issued for the project PID 34345 to use the SecureChatAI features.
-
Find your Python Bin Directory
which cappy
If it doesn’t show a path, try
which python3to find your Python bin directory. -
Add to Path For macOS (Zsh):
echo 'export PATH="/path/to/your/python/bin:$PATH"' >> ~/.zshrc source ~/.zshrc
Adjust accordingly for your shell/OS.
-
Verify Installation
cappy chat
This command should launch an interactive chat session if everything is configured correctly.
- Python 3.10+
- The
patchcommand (on Linux/macOS by default) .envfile with valid credentials for SecureChatAICAPPY.mdin your project directory (for context)- (Optional)
.cappyignoreto filter out certain files in scanning/search
With these in place, you can run Cappy’s agent with:
cappy agent "Your task here"or the interactive chat with:
cappy chatThat’s it! Enjoy using Cappy Code for secure agentic automation.