Terraform for agents: how we define our software factory in code

A software factory for AI agents can be defined entirely as code, similar to using Terraform for infrastructure. This approach uses a central configuration file to manage agent orchestration, model selection, secrets, and automations. Defining the agent pipeline as code enables version control, experimentation, and collaboration between humans and agents to improve the system. The structure includes a main factory definition, individual agent files specifying roles and permissions, and automation files that trigger agent workflows based on external events.