Skip to main content
pragma-os is currently in private beta. During this period, usage is free for beta participants. Pricing will be announced when we move to general availability. Contact [email protected] to request beta access.
pragma-os supports Google Cloud Platform (GCP) resources including Cloud Storage, CloudSQL, and Secret Manager. It also includes the Agno provider for AI agent infrastructure, a Kubernetes provider for container workloads, and built-in Pragma resources for secrets and file storage. AWS and Azure support is on the roadmap. See the Store for the full list of available resources.
pragma-os uses Clerk for user authentication. When you run pragma auth login, you authenticate through a secure browser flow. Your credentials are never stored in plaintext. Provider credentials (like GCP service accounts) are managed separately and securely within the platform. See Authentication for details.
When a resource operation fails, pragma-os retries it automatically. If retries are exhausted, the failure details are preserved for debugging. You can inspect failures, fix the root cause, and retry. Resources stuck in processing state are automatically detected and recovered. See the Error Recovery guide for the full workflow.
Dependents are detached, not cascade-deleted. pragma-os removes the dependency reference from each dependent’s config and sends them back to their provider for update. The provider decides whether the resource can continue without the deleted dependency — if it can’t, the resource transitions to failed. This prevents accidental cascade deletions while letting providers handle graceful degradation. See Deletion Behavior for details.
Yes. If a resource’s dependencies aren’t ready yet, pragma-os stores it in pending state and automatically processes it once all dependencies become available. You can define your entire infrastructure in a single YAML file and apply it at once — pragma-os resolves the dependency graph and processes resources in the correct order.
Define your agent in YAML with a model, instructions, and optional tools, then apply it. See the Quickstart for a step-by-step walkthrough.
A FieldReference references a single output value from another resource (like a connection URL or API key). A Dependency[T] gives the provider access to the entire resolved resource — its config, outputs, and metadata. Use FieldReference for simple value passing; use Dependency[T] when the provider needs the full resource context. See Reactive Dependencies for details.
Use the @path syntax in your YAML config to upload files inline:
provider: pragma
resource: file
name: my-document
config:
  content: "@./document.pdf"
  content_type: "application/pdf"
Or use the SDK’s upload_file() method for programmatic uploads. See the File Uploads guide.
Yes. pragma-os and Terraform solve different problems and can coexist. Use Terraform for static infrastructure that rarely changes (VPCs, IAM, etc.) and pragma-os for dynamic data and AI infrastructure with dependencies that need automatic propagation. See How We Compare for more details on when to use each tool.
Email [email protected] for assistance. Beta participants also have access to a private Slack channel for faster response times.
Key roadmap items include:
  • AWS and Azure providers — Expanding beyond GCP
  • Pre-built AI pipeline templates — Full end-to-end infrastructure templates
  • Team workspaces — Collaboration features for organizations
  • Additional AI providers — More LLM and vector database integrations
Contact [email protected] to share feature requests or learn more about upcoming releases.