I have been trying to get LLMs to handle real infra/DevOps tasks, and the biggest issue is that they don’t have reliable context. Infra is deeply interconnected, one misconfig can break a system.

So I built Paarvai. It connects to your cloud and IaC (read-only), builds a dependency graph, and gives LLMs the full picture of how everything is mapped and configured.

Unlike tools that make on-the-fly calls (like Amazon Q), Paarvai stores the whole state and all relationships upfront, so agents or tools like Cursor/Claude/GPT can do things like:

1. Understand dependencies (“What depends on this SQS queue?”) 2. Identify breakage (“Will disabling this Lambda break an API route?”) 3. Generate IaC with full context (“Write Terraform for this component using my current infra.”)

We just launched the MVP with AWS support.

I’m looking for early users, completely free. If you try it and want something added, I will personally integrate your feature requests.

Link: https://paarvai.app

Happy to share details or answer questions.