ClawAI in your editor
ClawAI Coding Agent for VS Code
See how the VS Code extension reaches every model on your ClawAI subscription, with routing, comparison, reviewable diffs, and history kept on the platform.
Every model on your ClawAI subscription, inside the editor you already use. The extension is a thin client: your account, your quotas, your provider credentials and your conversation history stay on the platform, so the same thread you started in the browser continues in VS Code.
What it does
- Every model, one subscriptionNine frontier families and your local open-weight models, reachable from the editor with no API keys to paste. Routing happens on the platform, so the editor never holds a provider credential.
- Automatic or manual routingLet the router pick the model for each message, or pin a conversation to a specific one. The choice is the same one the web app makes, because it is made in the same place.
- Compare and judge, in the editorSend one prompt to several models at once and read the answers side by side, with an optional judge pass — the same comparison workflow as the web app, against the code you have open.
- Preview before applyEdits arrive as a reviewable diff, not as a surprise write. Nothing touches your working tree until you accept it.
- Context you can inspectEvery answer carries a receipt: which files were read, which model answered, and what it cost against your allowance. When an answer is wrong, you can see what it was looking at.
- Concurrent conversationsSeveral titled chat tabs at once, two running concurrently against different models, with backend history restored in place.
What you need
VS Code 1.98 or later, and a ClawAI account. The extension connects to ClawAI’s hosted platform or to your own self-hosted deployment — you choose which at sign-in.
Questions people ask
- Do I need a separate subscription for the extension?
- No. The extension uses the ClawAI account you already have, and draws on the same allowance as the web app. There is nothing extra to buy.
- Does my code get sent to a model provider?
- Only what a request needs, and only to the model that answers it — the receipt on each answer names that model. Pin a conversation to a local open-weight model, or point the extension at a self-hosted deployment, and nothing reaches an external provider.
- Does it work with a self-hosted ClawAI?
- Yes. The extension asks for the backend URL at sign-in, so it works against ClawAI’s hosted platform or an instance running entirely on your own infrastructure.
- Can I keep using the web app as well?
- Yes, and the same conversations appear in both. History lives on the platform, so a thread started in the browser continues in the editor and back again.