Getting Started#

Atlas is distributed as a single standalone binary. You can install it using our automated scripts or via Go.

Installation#

curl -fsSL https://atlas.yashh56.me/install.sh | sh

Make sure your $(go env GOPATH)/bin directory is in your system's PATH.

Alternatively, to build from source:

git clone https://github.com/Yashh56/atlas
cd atlas
go build -o atlas ./cmd/atlas

Configuring LLM Providers#

Atlas requires an LLM API key to perform its "self-healing" build checks. You can provide this securely via the CLI. Atlas supports Anthropic, OpenAI, Mistral, Gemini, Groq, and xAI.

# Example: Store your Anthropic key securely in the OS keychain
atlas models set anthropic

You can verify your configuration at any time by running:

atlas models

Your First Deployment#

Atlas is designed to be completely zero-config. Simply navigate to your project directory and run the atlas command interactively.

cd ./my-web-app
atlas .

The interactive wizard will guide you through:

1. Model Selection: Choosing which LLM to use for fixing code.
2. Action Selection: Choosing whether to just build, or to build and deploy.
3. Provider Selection: Choosing a deployment provider (like Vercel, Render, or Netlify).
4. Authentication: If you aren't logged in, Atlas will prompt you to provide a token or automatically run the provider's CLI login.

Once configured, Atlas will detect your framework, validate your Git state, build your application, fix any build errors automatically, and finally deploy the project, returning your live URL!

Non-Interactive Mode (CI/CD)#

If you are running Atlas in a CI/CD pipeline (like GitHub Actions), you can bypass the interactive wizard by providing flags:

export ANTHROPIC_API_KEY=sk-ant-...
export VERCEL_TOKEN=...

atlas . --action deploy --provider vercel --model anthropic --auto-rollback-on-unhealthy