We're opening the doors. As of June 22, 2026, the DODIL AI cloud is live in alpha — and signing up is free.

DODIL is a vertically-integrated AI cloud running on our own hardware and network — sovereign infrastructure built for EMEA. One consistent, agent-addressable platform for the things AI workloads actually need: durable storage and memory, large-scale data processing next to that data, and a directory of models behind one API. If you're building agents, it's the backbone you offload to. (For the architecture, see Inside the DODIL AI Cloud.)

What's in the alpha

Every plane of the platform is available to build on today:

  • DataK3 — one bucket = objects (S3) plus one HTAP tables engine serving three pillars over the same rows: SQL (DuckDB dialect, Postgres-wire), Vector (pgvector), and Graph (Cypher/Bolt). No ETL, no second copy. Index a corpus for semantic search with a single recipe install.
  • Ignite — serverless functions and containers with ~200 ms cold starts, scale-to-zero, and deployed functions exposed as MCP tools.
  • Models — an OpenAI-compatible endpoint over 50+ hosted models plus bring-your-own: chat, embeddings, rerank, transcription. One key, one bill.
  • Scriptum — typed pipelines and recipes that wire the primitives together.

And everything is reachable two ways: the dodil CLI, or by prompting an agent — Claude Code, Cursor, VS Code, Codex — through the DODIL MCP.

Get started (free)

  1. Sign up at register.dodil.io — no card, no cost during alpha.
  2. Install the CLI and authenticate: dodil auth login — or connect the DODIL MCP to your agent.
  3. Build something — our tutorials take you from an empty bucket to a working RAG agent, a visual product search, an AI SIEM, and more. Full reference lives in the docs.

Your first workload is a few commands — or a single prompt:

"Create a DataK3 bucket, install the document-rag recipe, upload these PDFs, and answer questions over them."

What "alpha" means

Alpha is early on purpose. It's free, it's open, and it's real infrastructure you can build on today — but expect rough edges, and expect us to say so plainly (our tutorials already flag the gaps). Your data, embeddings, prompts, and model calls stay on our hardware, in-region. What we most want back is feedback: tell us what you're building and where it bites.

Welcome to the alpha. Let's build.

Start here: