wAIve, in your terminal.
wAIve CLI is the wAIve coding agent for developers. Point it at the models you run yourself or any OpenAI-compatible server. On your own GPUs: no API bill, no code leaving the building, no rate limit but your hardware.
Early access
Any OpenAI-compatible server
Local servers like vLLM, Ollama, LM Studio and llama.cpp, or a hosted provider. If it speaks the OpenAI API, wAIve CLI can use it.
Local first
waive setup finds and benchmarks every model server on your network.
Helpers in worktrees
Sub-agents work in parallel, each in its own git worktree.
Checkpoints
Every step is checkpointed. waive undo takes it back.
Autonomy dial
Plan, manual, normal or auto. You choose how much it does alone.
Goal loops
Keep going until a real check passes: --until "pytest -q". If one model server drops, the next in the chain picks up.
Add a server in one line
Give it the URL and the name of the environment variable that holds the key. wAIve CLI probes the endpoint for tool calling and context size before it uses it, so you find out right away whether a model can do agent work.
waive setup
waive providers add mine --url <your-url> --key-env MY_API_KEY
You pass the name of an environment variable, not the key itself.
Proof
Every example in the docs is executed by a test. Works with your existing CLAUDE.md / AGENTS.md files.
Honest limits
- Linux first. Windows runs it through WSL.
- Your model needs tool calling and at least a 32K context window. The probe tells you if it doesn't.
Every mTec hardware box ships with wAIve CLI installed.
Agentic coding for your dev team, pointed at the box, with no code leaving the building.
Start with a free exploration call.
Tell us what you're trying to do and what's in the way. If AI fits, we'll show you where. If it doesn't, we'll say that too.