The open-source agentic app builder.
Think Lovable, Replit, or v0, but self-hostable and self-contained.
Build at /tinykit, deploy to /
Tinykit is an open-source platform for building and deploying web apps with AI. It manages code, database, content, and deployment under the hood, letting you focus on the details.
- Agentic Builder: Prompt the AI to write code, create fields and tables, and wire it all up.
- Self-Hosted: Your server, your data. Powered by PocketBase. Docker optional.
- Realtime Database: Auto-generated tables that sync in real-time.
- Code Editor: Direct access to your source. One Svelte file per app.
- Content Fields: Edit text without touching code—like a mini CMS.
- Design System: Update colors, fonts, shadows from a visual editor.
- Time Travel: Snapshots on every change. Undo anything.
- Image Uploads: Built-in asset storage for images and files.
- Bring Your Own LLM: OpenAI, Anthropic, Gemini, DeepSeek, or a compatible local/custom server.
- Use Without AI: Start from a template, edit code and content, and run your apps without an LLM API key.
Run hundreds of apps on one server. Point any domain → get a working app.
Tinykit is early alpha. This maintenance update improves Docker setup, custom AI endpoints, domain setup, Canvas uploads, and preview/publishing reliability. See the changelog for changes and upgrade instructions. Feedback and bug reports are welcome on GitHub and Discord.
The easiest way to get started is with Railway. Otherwise, you can run it locally with Docker or Node.js.
Deploy on Railway and complete the setup screen. Configure AI from Settings, or start with templates and manual editing.
Or deploy with Docker or a VPS
The simplest way to run Tinykit locally:
git clone https://github.com/tinykit-studio/tinykit.git
cd tinykit/deploy/docker
docker compose up -d --buildThis starts Tinykit on port 3000. Visit http://localhost:3000/setup to create your account.
See deploy/docker/README.md for more options.
Use Node.js 22 or newer and npm:
# Clone and install
git clone https://github.com/tinykit-studio/tinykit.git
cd tinykit
npm install
# Start dev server (PocketBase was downloaded during installation)
npm run devVisit http://localhost:5173/setup to complete setup.
Open /tinykit/settings to choose a provider and enter its exact model ID. For a local server, choose Custom / local (OpenAI-compatible) and set the API base URL, for example http://localhost:1234/v1 for LM Studio. An API key is optional when your server does not require one. Select a model that supports tool calling.
Custom endpoints use Chat Completions for both the builder and history summaries. Test checks connectivity and authentication; it does not verify generation or tool support. Provider setup and Docker networking explain DeepSeek, local models, and environment configuration.
Templates, content editing, and published apps can work without AI. This release connects to model APIs; local coding-agent and subscription-based integrations are not included.
Each app gets its own domain. Point multiple domains to one Tinykit server:
calculator.myserver.com/ → Serves calculator app
calculator.myserver.com/tinykit/studio → Edit calculator app
blog.myserver.com/ → Serves blog app
blog.myserver.com/tinykit/studio → Edit blog app
| URL | Description |
|---|---|
/ |
Your production app |
/tinykit/studio |
Edit current domain's app |
/tinykit |
See all apps |
/tinykit/settings |
LLM configuration |
Starter templates to kickstart your apps:
| Category | Templates |
|---|---|
| Productivity | Kanban, Notes, Canvas, Timer |
| Finance | Expense tracker, Invoice generator |
| Content | Bookmarks, Recipes |
| Social | Linktree, Poll, Event RSVP |
| News | HN reader, RSS reader |
--
Build your digital homestead.
--
Building a website? check out Pala CMS