Local first
It finds LM Studio, Ollama, vLLM, llama.cpp, Jan and Lemonade on your machine by itself. With a local model, your code never leaves the computer.
Reges Code puts local and free models first, switches to the next free model when one is rate-limited, lets a team of models review each other's work, and tells you plainly when a model only said it did something.
Press CtrlK to try the palette
It finds LM Studio, Ollama, vLLM, llama.cpp, Jan and Lemonade on your machine by itself. With a local model, your code never leaves the computer.
When a free model is rate-limited or down, it moves to the next free one (the Reges gateway, other free models, then your local ones) and tells you. It never switches you to a paid model on its own.
If a turn claims "files created, ready to run" but wrote no files and ran no commands, Reges Code puts a clear warning under the reply. No more believing work that never happened.
Give roles to different models: architect, coder, reviewer, security, judge. They hand work to each other in a pipeline, work in parallel with a judge, or debate for a few rounds. Load and unload local models from the chat. Free and local models by default.
Local servers are detected on their own. Cloud routers work with your own key. MCP servers come preinstalled, and extensions come from Open VSX.
Reges Build turns a prompt into a full-stack app you own. Then ship it where you want. No credit meter ticking while the AI fixes its own mistakes.
Free while in beta. Windows 10/11, 64-bit. It updates itself, so you download once. macOS and Linux are next.
Download for WindowsWindows may show a SmartScreen notice for this beta. Choose More info, then Run anyway.