Getting Started

Follow these fundamental steps to start using STAN.

1. Set up a STAN agent

The STAN workflow requires an assistant configured with the "bootloader" prompt, which allows it to read your archive.tar and load the project's system prompt securely.

For the best experience (TypingMind)

We recommend TypingMind for its superior support of long, stable system prompts and convenient UI features.

1. Get TypingMind: Purchase a license (Extended required, Premium recommended) at typingmind.com.
2. Configure API Keys: Set up your OpenAI or Google Gemini API keys in TypingMind settings.
3. Import a STAN Agent:
   • GPT-5.2: Import Agent
   • Gemini 3 Pro: Import Agent

ChatGPT Users

• Custom GPT: Use the STAN Custom GPT.
• Manual Setup: Create a new Project and copy the contents of the bootloader prompt into the project instructions.

Google Gemini Users

• Custom Gem: Use the STAN Gem.
  • Note: Code blocks in the Gemini web app can be difficult to copy correctly; TypingMind is preferred for Gemini models.

---

2. Install & run STAN

Prepare your repository for AI-assisted development.

Install

Install the CLI globally:

npm i -g @karmaniverous/stan-cli
Copy

Initialize

In your repository root, initialize STAN. During initialization, you will be prompted to select key scripts (like test, lint, typecheck, build) to include in the context. See Configuration for details.

stan init
Copy

Run the loop

1. Run: Execute stan run to build/test and generate context. Commit your changes (save the state).
2. Snap: Execute stan snap to baseline diffs.
3. Share: Drop the artifacts into chat:
   • New thread: archive.tar + outputs.
   • Existing thread: archive.diff.tar + outputs.
4. Patch: Discuss with the agent. It will generate patches. Apply them with stan patch.

Return to Run to verify.

See The STAN Loop for details on the development cycle.