LaunchKit AI is an open-source AI app builder that turns prompts into runnable web apps with sandboxed coding agents. It combines a chat-first product surface with OpenAI-powered agent workflows, Inngest orchestration, E2B code sandboxes, tRPC APIs, Clerk authentication, Prisma, and Postgres.
The project is designed for maintainers and builders who want a practical reference app for prompt-to-app generation, safe sandbox execution, usage limits, and previewable artifacts.
Most AI app builders hide the interesting parts. LaunchKit AI keeps the architecture visible so open-source maintainers can inspect, fork, and adapt the core loop:
flowchart LR
A["User prompt"] --> B["tRPC protected mutation"]
B --> C["Inngest code-agent/run event"]
C --> D["OpenAI agent network"]
D --> E["E2B sandbox tools"]
E --> F["Generated files + preview URL"]
F --> G["Prisma project fragment"]
G --> H["Next.js UI"]
- Sandboxed coding-agent workflow with
@inngest/agent-kitand E2B. - OpenAI model calls for code generation, fragment titles, and user-facing responses.
- Next.js App Router, React 19, Tailwind CSS, and shadcn-style UI primitives.
- tRPC routers for projects, messages, and usage state.
- Clerk authentication with protected project routes.
- Prisma schema for projects, messages, fragments, and usage accounting.
- Rate-limited free/pro usage model through
rate-limiter-flexible. - CI-ready lint, typecheck, and build scripts.
See docs/demo.md for smoke-test verification and the full provider-backed demo checklist.
| Area | Tools |
|---|---|
| App | Next.js 15, React 19, TypeScript |
| UI | Tailwind CSS 4, Radix UI, lucide-react |
| API | tRPC, TanStack Query |
| Agents | OpenAI, Inngest Agent Kit |
| Sandboxes | E2B Code Interpreter |
| Auth | Clerk |
| Data | Prisma, Postgres |
| Background jobs | Inngest |
git clone https://github.com/AndyY-Q/launchkit-ai.git launchkit-ai
cd launchkit-ai
npm install
cp .env.example .env.local
npm run devOpen http://localhost:3000.
Create .env.local from .env.example.
| Variable | Required | Purpose |
|---|---|---|
DATABASE_URL |
Yes | Postgres connection string used by Prisma. |
NEXT_PUBLIC_CLERK_PUBLISHABLE_KEY |
Yes | Public Clerk browser key. CI can use the documented dummy test key for build-only checks. |
CLERK_SECRET_KEY |
Yes for auth runtime | Server-side Clerk key. |
OPENAI_API_KEY |
Yes for agent runtime | OpenAI key used by @inngest/agent-kit. |
E2B_API_KEY |
Yes for sandbox runtime | E2B key used to create sandboxes. |
INNGEST_EVENT_KEY |
Yes for deployed jobs | Inngest event key. |
INNGEST_SIGNING_KEY |
Yes for deployed jobs | Inngest signing key. |
NEXT_PUBLIC_APP_URL |
Recommended | Canonical app URL for deployed environments. |
- Create a Clerk application.
- Copy the publishable and secret keys into
.env.local. - Configure sign-in and sign-up URLs for
/sign-inand/sign-up.
For CI build verification only, this repository uses a non-secret dummy publishable key:
NEXT_PUBLIC_CLERK_PUBLISHABLE_KEY=pk_test_ZHVtbXkuY2xlcmsuYWNjb3VudHMuZGV2JA
CLERK_SECRET_KEY=sk_test_ZHVtbXkThese values are intentionally not valid for real runtime authentication. When these exact dummy values are present, LaunchKit AI enters smoke-test mode: Clerk middleware, ClerkJS, the pricing table, and authenticated project lists are skipped so CI can verify the public app shell without contacting Clerk.
Create an API key and set OPENAI_API_KEY. The current agent configuration lives in src/inngest/functions.ts.
Create an E2B account and set E2B_API_KEY. The sandbox template is in sandbox-templates/nextjs.
Run Inngest locally when testing background jobs:
npx inngest-cli@latest devThe app exposes the Inngest handler at /api/inngest.
Run migrations after configuring DATABASE_URL:
npx prisma migrate devnpm run dev
npm run lint
npm run typecheck
npm run buildThe npm run ci command runs all three verification steps.
Before applying to OpenAI Codex for OSS, make sure GitHub can review the project without authentication:
- Set the repository visibility to public.
- Confirm
https://github.com/AndyY-Q/launchkit-ailoads in a private browser window. - Confirm the repository appears on the public
AndyY-Qprofile. - Confirm the GitHub API returns metadata from
https://api.github.com/repos/AndyY-Q/launchkit-ai. - Add repository topics such as
openai,coding-agent,e2b,inngest,nextjs,trpc,oss. - Add a short repository description:
Open-source AI app builder for turning prompts into runnable web apps with sandboxed coding agents.
LaunchKit AI should be presented as an open-source AI app builder with a transparent sandboxed coding-agent architecture. The strongest application angle is:
LaunchKit AI gives open-source maintainers and builders a working reference for prompt-to-app generation, isolated code execution, artifact previews, and usage governance using widely adopted web primitives.
Contributions are welcome. Start with issues labeled good first issue or help wanted, and open a pull request with a clear description, screenshots for UI changes, and the output of:
npm run ciSee ROADMAP.md.
MIT. See LICENSE.