Dashboard
Launch work, inbox items, approvals and projects stay visible.
Maravilla Studio
The AI-native application platform.
Maravilla Studio is where websites, business applications and AI agents share the same structured data and publishing workflows.
Studio schemaeditors · rules · tools
AI agentsMCP tools · MCP-UI
RaisinDB graphpages · apps · workflows
Branch publishingdraft · review · live
System mountsmail · files · calendars
For years, teams were asked to choose.
Pages over here. Business apps over there. AI agents waiting outside the workflow.
CMS Internal tools Workflow engine Vector index Agent UI
Studio starts somewhere else.
One structured model for the website, the business app, the workflow and the AI agent.
RaisinDB graph Studio schema Studio Flow Branch publishing MCP tools
The product surface
One place where people and agents run the work together.
Studio does not split human workflows from agent workflows. A task can belong to an editor, a reviewer, Studio's own agents or an agent your team builds. The handoff, permissions, data and publishing trail stay the same.
Content-driven apps
Every workspace can become an operating surface for a team.
Human + agent workflows
Send work to a person or an AI agent. Studio treats both as responsible actors.
Studio Flow
Approvals, tasks and durable workflows route to humans or agents without a separate system.
Studio Flow
Any signal can start the workflow.
A user can ask in chat, but Flow can also start from triggers, inbox activity, mounted-system changes or your own application calls. Agents can propose plans, wait for approval, call tools, hand work to people or other agents, and record every step against the same RaisinDB nodes.
Inputs
A chat, trigger, inbox event, API call or mounted-system change can start the work.
Plan
Agent intent can become visible, reviewable and approvable before it runs.
Flow
Approved work keeps running in the background with tasks, deadlines and audit.
Actors
A step can go to a reviewer, Studio AI or a customer-built agent.
Dashboard
Launch operations at a glance.
See what changed, what is overdue and what needs approval.
Content-driven apps
Every workspace can become an app.
Define the schema, editor layout, workspace and app entry in a Studio package. The app appears in Studio with navigation, list, grid or board views, record editing, permissions and actions without writing a separate admin surface.
Or describe the app in plain language and let Studio Builder build it. See how Builder works
Stories
Pages, variations and publish state.
Assets
Media with usage and dependency lookup.
Branding
Design tokens as editable content.
Sites
Blueprints, linked copies and locale overlays.
Projects
Briefs, deliverables and approvals.
MCP apps
Typed agent tools and inline MCP-UI widgets.
Studio Builder
A coding agent for your content platform.
Studio Builder works the way a coding agent works on a codebase: it reads what exists, plans, builds, tests and repairs. Except its codebase is your Studio: record types, apps, automations, AI agents and functions.
Describe what you need in plain language. Builder designs it, shows you the exact changes before anything is written, builds it, and proves each part works. It writes real code where exact logic is needed, and it never touches anything outside the pieces you let it change.
Stop, steer, pause and resume at any time. A build survives a restart and never writes the same change twice.
You approve one specific set of changes. If the build is revised, Builder asks again.
A part is done only when a check proves it. A later edit marks that proof stale.
What cannot be built yet is recorded, not improvised.
Stories
Pages, policies and variations live together.
Stories carry the structure, rules and publish state the renderer needs.
Inline editing
Edit the real site, inside Studio.
The live preview is editable because the renderer and Studio share structure.
Assets
Media knows where it is used.
Assets show publish state, focal points and every dependency.
Brand systems
Design tokens are content.
Colors, type, logos and spacing become reusable, governed data.
Sites
Blueprints for every market and language.
Start from a proven site, roll out a new edition and keep shared pages connected while local teams override what needs to change.
Blueprints
Create from parameters instead of starting from a blank page.
Linked sites
Copy from a source tree, inherit updates and break local pages when needed.
Translation
Auto-translate editable fields into locale overlays, then review them in Studio.
Projects
Production belongs in Studio.
Briefs, deliverables, owners and channels move through the same system that publishes them.
Production board
Readable status, visible work.
Project work moves through Studio as real deliverables, tasks and approvals.
Connectors
Mount external systems into the same graph.
RaisinDB gives Studio a connector framework for mailboxes, drives, calendars and internal systems. Use shipped adapters or build your own, then mount external data anywhere in a workspace so pages, approvals, tasks, custom apps and registered agents work on the same native nodes.
Microsoft 365, Google Drive, Google Calendar, IMAP and integration support can be enabled from Studio.
Choose workspace, branch, path, hierarchy and sync rules so external systems land exactly where teams work.
Customer connectors can map EHRs, portals and proprietary systems into RaisinDB without leaving the publishing model.
Microsoft 365 Mail read · delta · webhooks
Google Drive Files browse · sync
Calendar Events read · write
Custom EHR adapter customer-built
mount /hospital/admin-inbox mapped into RaisinDB
Studio content tree native nodes
Hospital Admin Inbox
2026
08
intake@riverside-hospital.test
raisin:Mail New referral ready for admission review
lab-results@riverside-hospital.test
raisin:Mail Results received for patient record
insurance@payer-network.test
raisin:Mail Preauthorization approved for scheduled care
Model data labeling
Your documents and images become training data.
Studio turns the material already in the graph into supervised training sets. Documents are extracted, chunked, formatted, curated and mixed; images are labeled with AI-assisted segmentation and classification. Both happen in the same workspace, so the dataset behind a model carries the same review, versioning and audit trail as everything else.
SourcesExtractChunkFormatCurateMixTrain
Recipes are content
Extraction, chunking and formatting rules live in Studio. Change a rule and the sample preview updates beside it.
Curate before you train
Filter, split and keep what belongs in the set. Every build is an immutable snapshot, so an old run stays reproducible.
Mix and compile
Blend builds with weighted shares, read composition, repeats and sample lengths, then compile the run.
AI-assisted labeling
Image segmentation and classification are proposed by a model. You accept, correct or redraw the mask instead of tracing it by hand.
Why teams switch
The Shift.
Studio is built for the next generation of applications: products where people and AI agents work on the same data. RaisinDB gives Studio one AI-native engine for content, applications, workflows and agents, combining hierarchy, graph, SQL, search, vectors, functions and triggers so every surface shares the same live state.
Legacy enterprise suites Maravilla Studio
Separate systems Content, app state, search and automation live apart.
Teams stitch product state across repositories, services, indices and workflow tools.
AI-native engine Hierarchy, graph, SQL, search, vectors, functions and triggers in one system.
Studio runs content, application state, workflow state and agent context on RaisinDB.
Sync and middleware External data is copied through APIs, jobs and adapter services.
Every integration creates another place where state can lag, drift or break.
Native mounts External systems become live nodes in the graph.
M365, Drive, calendars and customer systems appear where teams already work.
Bolted-on AI Agents work through derived context and custom tool layers.
Separate indices and wrappers make agent actions harder to approve, audit and trust.
Agent-ready execution The same schema defines editors, validation, MCP tools and MCP-UI widgets.
People and agents use the same permissions, rules, context and publishing controls.
Split product surfaces Websites, internal tools and workflows need separate stacks.
Teams rebuild the same models, access rules and UI patterns in different products.
Application workspaces Every workspace can become a website, dashboard, inbox or product app.
Pages, operational tools and customer apps run against the same live graph.
Fragmented release flow Content, data, app changes and agent actions are reviewed separately.
Approvals happen in different tools, leaving teams without one reliable audit trail.
Shared release flow Humans and agents draft, review, approve, publish and audit the same work.
Branch-based publishing keeps content, app state, workflow state and agent actions together.
MCP apps
Build agent apps. Or use Studio's.
Studio can expose your product as typed MCP tools with MCP-UI widgets. Use Studio's own agent workflow, connect ChatGPT or Claude, or build your own agents. They enter the same workspace as users, with the same permissions, tasks and publishing confirmations.
Declare the agent actions and MCP-UI widgets your product needs. Studio handles access, context and delivery.
Let an agent browse stories, inspect pages, edit fields, review dependencies and publish through the same workflow as a human.
Today 12:42 PM
Architecture
Schema to editor to renderer.
RaisinDB graph
Structured nodes, auth, branches, functions and MCP tools.
Studio schema
Types, policies, editors, apps and brand kits.
Studio surface
Inline editing, assets, projects, inbox and approvals.
Renderer
Websites and apps that read the same Studio graph.
Publishing
Edit on one branch. Publish from another.
mainAuthors edit, agents validate, workflows wait.
publishThe live site reads the approved branch.
The category shift
One model. Many products. Human editors, AI agents and MCP apps in the same workflow.
Maravilla Studio
Build around structured data.
Built on RaisinDB Explore the AI-native database