Deploy and Host Bolt.diy on Railway

Bolt.diy AI app builder interface

Bolt.diy is the open-source, self-hostable version of Bolt.new — an AI full-stack app builder that writes, runs, and previews web apps live in your browser. Describe what you want and it generates the code, installs the dependencies, runs it in an in-browser WebContainer, and shows you the result — then you edit by chatting. Bring your own model: OpenAI, Anthropic Claude, Google Gemini, Groq, Mistral, Ollama, and more.

Hosted AI app builders meter you on message credits. Bolt.new, Lovable, and v0 all charge per generation and cap you by plan. Bolt.diy on Railway uses your API key — pay only for the tokens you actually use, with no per-message credits and no monthly builder subscription.

What This Template Deploys

Service	Purpose
Bolt.diy	The full app — chat UI, code editor, in-browser WebContainer runtime, and live preview, served on the public web service

Single-service deployment. Code generation and execution happen client-side in the browser's WebContainer, so there's no separate build server to run. Add a persistent volume if you want generated projects and chat history to survive redeploys.

About Hosting Bolt.diy

Bolt.diy (the community fork of StackBlitz's Bolt.new) lets you use any LLM instead of being locked to one provider, and self-hosting means your prompts, generated code, and API keys stay on infrastructure you control. Running it means serving the app, wiring an API key for each provider you want available, and giving it a public HTTPS endpoint.

Railway handles the hosting and HTTPS automatically. Add your provider API keys as Railway Variables — only the providers you configure appear in the model picker — and the app is live at your Railway domain.

Typical cost: ~$5–10/month on Railway's Hobby plan for the app itself, plus whatever your chosen LLM charges per token. Groq and Gemini offer free API tiers, and local Ollama is free — so your per-generation model cost can be zero while hosted builders keep charging credits.

Deploy in Under 5 Minutes

Click Deploy on Railway — Bolt.diy builds automatically (~2–3 minutes)
Add an API key for at least one provider in Railway Variables (e.g. ANTHROPIC_API_KEY, OPENAI_API_KEY, GROQ_API_KEY, or GOOGLE_GENERATIVE_AI_API_KEY)
Open your Railway-assigned URL
Pick a configured provider and model from the picker
Describe the app you want — Bolt.diy generates, runs, and previews it in the browser

No local Node setup. No cloning. No credit meter.

Common Use Cases

Self-hosted alternative to Bolt.new — the same in-browser AI app builder without StackBlitz's message-credit metering; use your own API key and pay only for tokens
Self-hosted alternative to Lovable and v0 — build and preview full-stack apps by chatting, on your own infrastructure, with any model instead of a fixed provider
Multi-model AI coding workspace — switch between Claude, GPT-4o, Gemini, Groq, and local Ollama models for the same project without changing tools
Private prototyping for client work — prompts, generated code, and keys stay on your instance, not a third-party builder's servers
Learning and rapid MVP building — spin up working React, Vite, and Node prototypes from a prompt, edit them live, and export the code
Team app-builder on a shared instance — one deployment, one place to manage provider keys, accessible to your whole team from a browser

Configuration

Variable	Required	Description
`ANTHROPIC_API_KEY`	One provider required	Enables Anthropic Claude models
`OPENAI_API_KEY`	Alternative	Enables OpenAI GPT models
`GOOGLE_GENERATIVE_AI_API_KEY`	Alternative	Enables Google Gemini models
`GROQ_API_KEY`	Alternative	Enables Groq-hosted models (fast, free tier)
`MISTRAL_API_KEY`	Alternative	Enables Mistral models
`OLLAMA_API_BASE_URL`	Optional	Point at an Ollama endpoint for fully local models
`OPENAI_LIKE_API_BASE_URL`	Optional	Any OpenAI-compatible endpoint (OpenRouter, LM Studio)
`PORT`	Auto-set	Railway injects the port automatically

Only the providers you configure appear in the model picker. You need at least one API key — everything else is optional. Add a persistent volume if you want generated projects to survive redeploys; otherwise the workspace is per-session.

Bolt.diy vs. AI App Builders

	Bolt.diy (Railway)	Bolt.new	Lovable	v0
Pricing model	~$5–10/mo + your tokens	Message credits	Message credits	Message credits
Bring your own model	✅ Any provider	❌ Fixed	❌ Fixed	❌ Fixed
Local models (Ollama)	✅ Yes	❌ No	❌ No	❌ No
Self-hostable	✅ Yes	❌ No	❌ No	❌ No
Data / code privacy	✅ Your instance	❌ Their servers	❌ Their servers	❌ Their servers
Live in-browser preview	✅ WebContainer	✅ Yes	✅ Yes	✅ Yes
Open source	✅ MIT	❌ No	❌ No	❌ No

Dependencies for Bolt.diy Hosting

Railway account — Hobby plan (~$5–10/month) for the app
At least one LLM API key — Groq and Gemini have free tiers; Ollama is free and local
Optional: a persistent volume for generated projects and chat history

Deployment Dependencies

Bolt.diy GitHub Repository — source and releases
Bolt.diy Documentation — setup and provider config
Groq Console — free-tier API keys
Railway Volumes Documentation — persistent storage

Implementation Details

This template deploys Bolt.diy as a single web service on Railway with automatic HTTPS. Code generation and execution run client-side in the browser's WebContainer, so there is no separate execution backend to manage. Provider API keys are read from Railway Variables at runtime — only configured providers appear in the model picker.

Because the WebContainer runs in the visitor's browser, use a modern Chromium-based browser for best compatibility. Add a Railway volume if you want generated projects and chat history to persist across redeploys; without one, each session starts fresh.

Frequently Asked Questions

How is Bolt.diy different from Bolt.new? Bolt.diy is the open-source, self-hostable community fork of StackBlitz's Bolt.new. Bolt.new is hosted and meters you on message credits with a fixed model. Bolt.diy runs on your own infrastructure, works with any LLM provider (including local Ollama), and charges nothing beyond your own token usage.

Do I need an OpenAI key? No — any one provider works: Anthropic, OpenAI, Gemini, Groq, Mistral, or a local Ollama / OpenAI-compatible endpoint. Groq and Gemini have free tiers, and Ollama is free, so you can run Bolt.diy with zero per-generation cost.

Why does the model picker only show some providers? Bolt.diy only lists providers you've configured with an API key. Add the relevant key in Railway Variables (e.g. ANTHROPIC_API_KEY) and redeploy — the provider then appears in the picker. A blank or missing provider almost always means the key isn't set.

Do generated projects survive a redeploy? Only if you add a persistent volume. By default the WebContainer workspace is per-session. Mount a Railway volume to keep generated projects and chat history across redeploys.

Which browser should I use? Bolt.diy runs its code execution in an in-browser WebContainer, which works best in modern Chromium-based browsers (Chrome, Edge, Brave). Older or restricted browsers may fail to boot the WebContainer.

Is my code and data private? Yes. Self-hosted on Railway, your prompts, generated code, and API keys stay on your instance. The model provider you choose sees the prompts you send it, but nothing routes through a third-party builder platform.

Why Deploy and Host Bolt.diy on Railway?

Railway is a singular platform to deploy your infrastructure stack. Railway will host your infrastructure so you don't have to deal with configuration, while allowing you to vertically and horizontally scale it.

By deploying Bolt.diy on Railway, you get an open-source AI full-stack app builder — live in-browser preview, any LLM provider, and full code privacy — at ~$5–10/month plus your own token usage, with no message credits and no builder subscription.