The resource types

Everything durable that Ronja builds is a resource, and resources live inside features. You rarely create them through forms — the pattern across the product is ask Ronja: describe what you need in an exploration, and the agent creates the resource, with approval gates where the change affects others.

Here are the nine feature resources, plus an organization-level building block — the managed database — that sits alongside them. For a lookup table of scopes, creators, and approvers, see the resource types reference.

Tables

The data itself. Tables come in four kinds — Foundation, Integration, Derived, and Dynamic — depending on where their data comes from, and they carry builds, lineage, and version history. Tables are the input and output of almost everything else on this page. See Tables.

Files

Files enter Ronja two ways. You upload them in chat — data files like CSV and Excel become tables, while images, documents, and text files ride along as context for the agent. Workflows also produce files: reports and exports that appear on a feature’s File outputs tab, grouped by the workflow that made them. See Upload files.

Workflows

A workflow is a reusable job — a saved piece of work Ronja can run again and again, reading input tables and writing output tables or files. On a table’s page, the workflows that write it appear as its “Producers”. Workflows belong to a workspace and are the workhorse behind most recurring output. See Workflows and automations.

Notes

Notes are how you teach the agent, and they come in two kinds: Skills teach Ronja a task or procedure (“here’s how we calculate churn”), while Knowledge notes hold reference information about a feature’s data. Both make every future conversation smarter. See How Ronja learns.

Data apps

A data app is an interactive application Ronja builds on top of your data — something a colleague can open and use without ever touching a conversation. Data apps have a draft-to-published lifecycle with an archived state, and can be opened standalone in their own page. See Build a data app.

Automations

An automation pairs a trigger with an action: on a schedule or when a watched table rebuilds, run a workflow or a Saved Agent; on an inbound email or webhook, run a Saved Agent. Automations are how work keeps happening when nobody is in a conversation; each one can be paused and re-enabled, and its runs are tracked. See Schedule an automation.

Secrets

A secret is a stored credential — an API key or database login — that other resources use without ever revealing. Ronja can authenticate against your systems on the strength of a secret, but the value itself is never shown back to anyone, and it’s never included when a feature is exported. See Manage secrets.

MCP servers

An MCP server connects Ronja to an external tool system, adding new capabilities the agent can use in conversation — each server exposes a set of tools. There’s no setup form: you ask Ronja in chat to connect one. See Connect MCP servers.

Saved Agents

A Saved Agent packages a repeatable agent task: a reusable prompt, a model tier (Fast / Standard / Mega), and an explicit set of tool and resource grants. Once saved, it can be invoked from other conversations or run by automations — your best prompts, made into infrastructure. Saved Agents are created and edited in chat, not through a form. See Create a saved agent.

Managed databases

Unlike the resources above, a managed database belongs to your organization rather than a feature. It’s a real Postgres database Ronja provisions and runs for you — a live system you can build on (a CRM, an ops queue, application state), not just a place to analyze data. You can also mirror one into a feature so its tables become data you can query. Because it belongs to your whole organization (not one feature), stays running for the long haul, and only admins manage it, it’s a related-but-distinct building block from the nine feature resources. See Managed databases.

Tip

Not sure which resource fits a job? Describe the outcome to Ronja in an exploration — choosing the right building block is part of what the agent does, especially in Build mode.