The WidelAI coding agent inside VS Code. Ask for a change in plain language and
WidelCode reads your project, edits files, runs commands and checks its work,
using any model on your WidelAI plan: Claude, GPT, Gemini and more.
Get started
Install the extension and open the WidelCode view in the activity bar.
Run WidelCode: Log In and approve the sign-in in your browser. You need a
WidelAI account: sign up at https://widelai.com.
Open a folder and type what you want done.
Features
Chat with an agent that works in your project. It reads files, searches,
edits and runs your build and tests, and shows each step as it goes.
You stay in control. Choose how much it may do without asking: ask
before everything, edit freely but ask before commands, or full auto. Every
edit can be opened as a diff.
Checkpoints. Each run is checkpointed first; WidelCode: Revert Last
Checkpoint undoes it, including changes made by commands.
Separate branch. In a git repository the agent works in its own worktree,
so your checkout is untouched until you apply the changes or open a pull
request.
Command sandbox. On macOS and Linux, commands can only write inside the
project.
Your project's setup. Reads AGENTS.md, CLAUDE.md, .widelai/rules,
skills, custom agents, hooks and MCP servers. Hooks and MCP servers run only
after you trust the project.
Cloud runs. Start a run on a GitHub repository from VS Code, follow it in
the Cloud runs view, and get a pull request when it finishes.
Credits you can see. Runs spend your WidelAI credits; set a per-run budget
in settings.
