A self-hosted boilerplate for building AI chatbots over your documents. No US data transfers. No JavaScript frameworks. Deploy to a €5/mo Hetzner VPS in minutes.
Most RAG solutions either send your data to the US, lock you into expensive SaaS, or require a full-stack JavaScript team to customize.
SaaS RAG tools route data through US servers. Your client data leaves the EU. Your DPO is unhappy. You need a Data Processing Agreement you'll never get.
Everything runs on your server. Ollama for local inference, ChromaDB for local vectors. Data never leaves your VPS. Mistral (French) as the default cloud provider when you need one.
Paid boilerplates like ChatRAG and StartKit.AI are locked into Supabase, Pinecone, and OpenAI. You can't self-host the vector store. You can't run inference locally. Your GDPR story falls apart at the architecture level.
EuroRAG is self-hosted top to bottom. ChromaDB runs on your machine. Ollama runs on your machine. No cloud dependency, no API keys required. Add Mistral or OpenAI when you choose to — with full data residency tracking.
Most RAG boilerplates are React + Node.js + Python + Docker orchestration. Your Python team now needs JavaScript expertise to customize the frontend.
One language. Python handles the backend, Jinja2 handles the templates, HTMX handles interactivity. If you know Python, you can customize everything. No npm install required.
Everything you need to build a production RAG chatbot, nothing you don't.
Data residency tracking per provider (LOCAL, EU, US). Deletion cascades. Audit logs. Strict EU mode blocks non-EU providers entirely. Not a checkbox — a design principle.
FastAPI + Jinja2 + HTMX. One language, one codebase, one container. No React, no Node.js, no webpack. Your Python team owns the entire stack.
Ollama for fully local inference. Mistral API for EU-hosted cloud. OpenAI/Groq when you need it. Switch providers by changing one env variable.
Expose your private documents as an MCP tool. Claude Desktop, LangGraph, CrewAI, or any MCP client can query your knowledge base without custom integration.
Upload PDFs, DOCX, TXT, Markdown. Sync from local folders, crawl websites, connect Nextcloud. All self-hosted, no monthly connector fees.
Real-time SSE streaming in the UI. Full OpenAI-compatible API at /v1/chat/completions so your existing tools just work.
Core/custom separation architecture. Modify templates, add routes, swap services — your changes survive when you pull updates from the core engine.
Designed for Hetzner CX22 (2 vCPU, 4GB RAM). Docker Compose up and you're running. No Kubernetes, no managed services, no surprise bills.
CSRF protection, rate limiting, path traversal prevention, session management, structured logging with request IDs. Audited and production-ready.
From purchase to running chatbot on your own server.
Purchase and download the complete codebase. Unzip it on your server or local machine.
Copy .env.example to .env. Set your model provider (Ollama for local, Mistral for EU cloud), choose your vector store, done.
Run docker compose up -d. The app, Ollama, and ChromaDB start together. Everything stays on your server.
Drop your PDFs into the admin UI, wait for indexing, and start asking questions. Connect Nextcloud or a web crawler for ongoing sync.
Customize templates, add routes, change the branding. It's your code — build the product your clients need.
Compared against other paid RAG boilerplates and the leading open-source alternative.
| EuroRAG | ChatRAG | StartKit.AI | AnythingLLM | |
|---|---|---|---|---|
| Type | Boilerplate | Boilerplate | Boilerplate | Full app |
| Stack | Python + HTMX | Next.js + React | Node.js + React | JS + Python |
| Fully self-hosted | Yes — everything local | No — requires Supabase | No — requires Pinecone | Yes |
| Local LLM inference | Ollama built-in | OpenAI only | OpenAI primary | Yes |
| GDPR data residency tracking | Built-in | None | None | Manual |
| MCP server | Built-in | No | No | No |
| Designed for customization | Core/custom separation | Fork and modify | Fork and modify | Full app — hard to rip apart |
| Runs on €5/mo VPS | Yes | Needs Supabase | Needs Pinecone | Yes |
| Price | From €49 | $199–269 | $199+ | Free / $40 |
AnythingLLM is excellent if you want a ready-to-use app. ChatRAG and StartKit.AI are great if you're building on Next.js/Node.js. EuroRAG is for Python developers who want a self-hosted, GDPR-compliant starting point they fully control.
No subscriptions. No per-seat fees. No "contact sales." Buy it, own it, ship it.
All tiers include full source code
EuroRAG gives you the technical infrastructure for GDPR compliance — data residency tracking, deletion flows, audit logs, consent management, and the ability to run everything locally or on EU servers. However, GDPR compliance also involves organizational measures (privacy policies, DPAs, etc.) that depend on your specific use case. We provide the tools; you'll still need legal review for your deployment.
Any model that runs on Ollama (Mistral, Llama 3, Qwen, Gemma, etc.) for fully local inference. Any OpenAI-compatible API (Mistral API, OpenAI, Groq, Together AI) for cloud. Or self-hosted inference servers like vLLM or LocalAI. Switch providers by changing environment variables — no code changes needed.
The Developer tier gives you the RAG engine, OpenAI-compatible API, and MCP server — without the chat UI or admin panel. It's for developers building agents or custom interfaces who want a GDPR-compliant private knowledge base they can query programmatically. Think of it as the backend for your own AI workflows.
Probably not. EuroRAG is source code — a boilerplate for developers and agencies building RAG products. You need to be comfortable with Python, Docker, and basic server administration. If you want a ready-to-use app, AnythingLLM might be a better fit.
React is powerful, but it means your team needs JavaScript expertise, a Node.js build pipeline, and a separate frontend deployment. EuroRAG uses Jinja2 templates + HTMX, which means the same Python developer who writes the API can also modify the UI. For a boilerplate you're meant to customize, this is a significant advantage.
ChatRAG ($199+) and StartKit.AI ($199+) are both excellent boilerplates, but they're built on Next.js/Node.js and depend on cloud services (Supabase, Pinecone, OpenAI). EuroRAG is fully self-hosted, runs entirely on your infrastructure with local inference via Ollama, and includes GDPR compliance tooling that neither competitor offers. If you're in the EU and care about data sovereignty, EuroRAG is the option that's GDPR-compliant by architecture.
Yes, with caveats. A Hetzner CX22 (2 vCPU, 4GB RAM) can run the app, ChromaDB, and small Ollama models for light usage (~10 concurrent users). For larger models or heavier loads, you'll want more RAM. The recommended setup is a CAX21 (4 ARM vCPU, 8GB) at €7.49/mo.
MCP (Model Context Protocol) is the open standard for connecting AI agents to data sources. With EuroRAG's MCP server, any MCP-compatible client — Claude Desktop, Cursor, LangGraph, CrewAI — can query your private documents without custom integration. Your documents stay on your server; agents just ask questions.
March 2026. Leave your email above and you'll get a single notification when it's available — no drip campaigns, no spam.
No, the software is not returnable. If you have more questions or would like to see code snippets - get in touch.
Get the source code. Deploy on your server. Own your AI stack.
Get Notified