RAGNA Studio is an AI workspace where agents are your co-workers. Your agents read your documents and keep their state in datasets. They delegate to each other and run on a schedule.
Core ideas
- Workspaces hold every resource: agents, datasets, documents, tasks, chats, workflows and media. Agents only see the workspace they run in.
- Agents are specialists. Each has its own instructions, model, tools and knowledge documents. It keeps a memory that it writes itself.
- Datasets are tables with typed columns. You edit them in a grid. Agents read and write them through tools, so a dataset works as a queue and as durable agent state.
- Workflows run on a visual canvas and a cron schedule. A team node lets a lead agent delegate to up to five specialists in parallel. Every run keeps a full trace.
Two ways to run it
| Hosted | Self-hosted | |
|---|---|---|
| Where | app.ragna.io | Your own server |
| Setup | Sign in and start | Docker Compose on one VM |
The self-hosted version is open source under AGPL-3.0. You host it, you own the data, and you bring your own model keys or run local models.
To try it on your own machine, follow the Quickstart. To run it for a team, see Self-hosting.
Supported providers
| Area | Providers |
|---|---|
| Chat and agents | Anthropic, OpenAI, Google (GenAI and Vertex), Mistral, LM Studio (local) |
| Embeddings | OpenAI |
| Images | FLUX (Black Forest Labs), OpenAI, Imagen |
| Video | FLUX 3 video, Veo |
| Web search | SerpAPI |
| Gmail, Outlook | |
| Sign-in | Google, Microsoft |
| Database | PostgreSQL 18 with pgvector |
| File storage | Any S3-compatible storage (Cloudflare R2, AWS S3, …) |
Privacy
RAGNA Studio sends no telemetry. A self-hosted instance only reaches the services you configure, such as your LLM provider, storage and mail account.