Docs menu
Coding with AI
Which AI coding tools are on the team devices, which models each may use, and the review every AI-written change needs.
Two AI coding tools are installed on the team devices: Claude Code and OpenCode. Both run in a terminal, in the folder of the project you are working on, and both read and edit files there. The commands on these pages are checked against each tool’s official documentation. The model rules below are team policy.
Model policy
| Tool | Allowed models | Who decides |
|---|---|---|
| Claude Code | Sonnet. Opus only with your team lead’s approval, given before you use it | Team policy |
| OpenCode | Free models only, because those are the only ones the team has access to | Team access |
Claude usage is a shared, limited allowance for the whole team. Rerunning the same prompt, or asking Claude what a search answers, uses it up, and heavy waste can get Claude access restricted for everyone. Match the tool to the size of the task: a quick question goes straight to the model, and a long multi-file change goes through the OpenCode Delegate plugin.
- Claude CodeStart Claude Code in a project, keep it on Sonnet, and hand long tasks to OpenCode through the delegate plugin.
- OpenCodeRun OpenCode interactively or for one question, pick one of its free models, and keep private material out of free-model sessions.
- Prompting and reviewPrompts that get reviewable changes from an AI tool, and the review and testing an AI-written change needs before it runs on a car.