Run coding agents in a secure sandbox
A simple SDK for safely running Codex and Claude in your app or workflow. Open-source and MIT licensed.
npm i @vibe-kit/sdk
Embed Codex or Claude Code in your app
import { VibeKit, VibeKitConfig } from "@vibe-kit/sdk";
const config: VibeKitConfig = { agent: { type: "codex", ... }, ... };
const vibeKit = new VibeKit(config);
const prompt = "Create a todo list";
const mode = "code";
const result = await vibeKit.generateCode({
prompt,
mode,
callbacks: {
onUpdate: (data) => console.log("Update:", data),
onError: (error) => console.log("Error:", error),
},
});
console.log("Result:", result);
Secure and Private
Runs code in a secure sandbox and private sandbox.
Customizable
Customize your environment, install packages and more.
Any Model
Choose any model to power Codex or Claude Code.
Streaming
Supports streaming the agent output to your UI.
Code or Ask
Supports code generation or asking questions.
Github Support
Work on your codebase, create PRs and more.
Supports your favourite sandbox provider
Supports E2B. Daytona, Modal, Fly.io coming soon.
Use any compatible sandbox runtime that meets your security and isolation needs.
Use cases
Coding agents are adaptable by design. Run them wherever they're useful, across tools, workflows and environments you control.
Internal Tools
Use coding agents to support debugging, refactoring, or automation inside secure engineering environments. Useful for onboarding helpers, CI bots, and internal dev tools.
App Features
Add code generation or code Q&A to your product. Enable users to scaffold apps, generate components, or explore code with natural input.
Prototyping Workflows
Use coding agents to scaffold features, generate boilerplate, or explore different implementation paths. Useful for iterating on ideas.
Integrations
Let agents handle common setup tasks like adding analytics, wiring up authentication, or configuring SDKs. Skip boilerplate and speed up implementation.