Hi everyone! We're a small infra team moving faster and I'm researching AI coding assistants to help with our Terraform and Ansible work. We need something that understands HCL and YAML deeply, can suggest modules, and help debug plan errors. Our main goal is consistency and reducing repetitive boilerplate.
I've been testing a few options with some real tasks from our repos:
* Generating a Terraform module for an AWS S3 bucket with proper logging and IAM policy.
* Converting an Ansible playbook from using the `yum` module to `dnf`.
* Explaining a tricky `terraform state mv` scenario.
So far, my impressions are:
- **GitHub Copilot**: Great for inline completions in VS Code, especially for repetitive Terraform resource blocks. Sometimes struggles with more complex Ansible logic or newer provider syntax.
- **Cursor**: Really impressive for larger refactors. The agent mode can handle "update the Terraform version constraint in all our modules" pretty well. Context handling is a strong point.
- **Claude (in Cursor)**: Excellent at explaining *why* a `terraform plan` might be failing. Gives clearer, more reasoning-based answers for Ansible role structure changes.
- **Tabnine**: Feels faster for single-line completions, but the suggestions for HCL can be less accurate than Copilot's.
For a team our size, I'm leaning towards Cursor for its project-wide awareness and the flexibility to use different models. The Claude integration seems to "get" infrastructure-as-code concepts better.
Has anyone else done a similar comparison specifically for IaC? I'm curious about cost vs. value for a 5-person team and any gotchas with onboarding.
– Julie
Data-driven decisions.