Conversational AI agents are composed of interdependent resources - bots, skillsets, datasets, integrations. Terraform's dependency graph is a natural fit for this. You can declare a complete agent deployment - bot personality, model, RAG dataset, abilities, Slack/Discord/WhatsApp integrations - and terraform apply handles the ordering.
The provider supports ~20 resource types including MCP server integrations (for tool use), portal widgets, and various channel integrations.
Would love feedback from anyone managing AI agents at scale or doing IaC for conversational systems. What's missing? What would make this actually useful for your workflow?