Adaptive model levels
Each new task starts at your chosen level, then can continue at Strong or Maximum when it needs more capacity.
Multi-model coding · Inside VS Code
CodeMux plans, codes, and reviews each task with the model suited to that step — OpenAI GPT, Anthropic Claude, DeepSeek, xAI Grok, or Google Gemini — then shows you the diff, the checks, and the cost in one place.
Free to install · $1 free credit · No subscription · Your API or ours
Fix checkout validation.
Add a regression test.
− if (!email) {+ if (!isValidEmail(email)) { return showError(); }Workflow illustration, not a recorded run. Models and checks vary by task.
One workflow, five providers
Why CodeMux
Use a different model where it helps, and keep every result in one place.
Each new task starts at your chosen level, then can continue at Strong or Maximum when it needs more capacity.
Every run shows the files it changed, the checks it ran, and what it could not verify — so you know exactly where to look before you accept.
Each run shows its total and which model handled which step. Managed models share one balance; your own API is billed by your provider.
How it works
Four steps, all inside VS Code 1.93 or later.
Need a hand? Contact supportInstall CodeMux from the Visual Studio Marketplace, then open your repository in VS Code.
Open the CodeMux sidebar and sign in with GitHub or Microsoft. New accounts get $1 in free credit to try a real task — add funds later, or connect a compatible API of your own.
Say what you want to happen and how to check it. Routing picks the models by default; you can override it.
Look over the changed files, the checks that ran, anything still unverified, and the cost — then decide whether to keep it.
Pricing
No subscription and no seat fee. Choose how you want to pay for model usage.
See pricing detailsTop up one CodeMux balance and use any model in the managed catalog. Start from $5.
Connect a compatible endpoint. Your provider bills you at its own rates; CodeMux charges nothing for these requests.
No. New work starts at Fast by default and can continue at Strong or Maximum when more capacity is needed. You can configure all three levels.
CodeMux runs your repository’s checks and records the results, including anything it could not verify. You review the changes and the evidence before accepting them.
The extension is free, and new accounts start with $1 in free credit to try a real task. Managed models then draw from a pay-as-you-go balance. If you connect your own API, your provider bills you directly and CodeMux charges nothing for those requests.
Open your repository, describe the change, and review the result. Free to install, nothing to subscribe to.