Skip to content

Capability availability

This page is the availability contract for the build guides.

It prevents examples from implying that every layer is operated directly by MIOSA today.

Status labels

Available now

MIOSA exposes and operates this capability through its current platform, CLI, SDK, or API.

Available through integrations

Your MIOSA application can use the capability through an external provider, API, or MCP connection.

Self-hostable now

You can deploy the application or service on current MIOSA compute, subject to its runtime and resource requirements.

Coming soon

The capability is directional and should not be treated as a production commitment until its product documentation marks it available.

Requires GPU

The workload requires accelerator capacity that may be supplied by an external provider today or by a future MIOSA runtime.

Activation gated

The implementation and packaging exist, but production activation is blocked until an explicit acceptance requirement passes.

Current product layers

LayerStatusWhat you can build
Isolated Linux sandboxesAvailable nowCoding agents, task workers, builds, scripts, document processing
Files, exec, streaming, and snapshotsAvailable nowStateful work, checkpoints, artifacts, debugging
Live port previewsAvailable nowAgent-built application review and development sharing
Persistent desktop computersAvailable nowBrowser agents, native applications, authenticated sessions
Managed Postgres, Redis, storage, and volumesAvailable nowProduct state, memory, queues, files, and caches
Immutable deployments and domainsAvailable nowPublic, private, and customer-facing applications
Workspace and external attributionAvailable nowWhite-label and downstream customer products
MCP and application toolsAvailable now or through integrationsBusiness systems, model providers, and custom APIs
OpenComputersAvailable nowConnect owned machines to MIOSA orchestration
AWS customer-cloud foundationsActivation gatedPrivate VPC, identity, host templates, and acceptance packaging
GCP customer-cloud foundationsActivation gatedPrivate network, service account, instance template, and acceptance packaging
Model APIsThrough integrationsImage, video, audio, speech, embeddings, and LLM products
Open-source AI interfacesSelf-hostable nowPrivate AI workspaces connected to external inference
Managed GPU model executionComing soonPrivate model endpoints and accelerator-backed generation
Built-in managed agent orchestrationComing soonNative responsibility loops and managed agent state

Available now

You can build and deploy the entire web application, control plane, customer portal, database layer, tool layer, and background execution path today.

You can also provide persistent computers and isolated sandboxes to your users through your own product.

Available through integrations

Use integrations when another provider owns a specialized real-time or hardware-intensive layer.

Common examples include:

  • Image and video generation.
  • Speech recognition and voice synthesis.
  • Telephone media transport.
  • Foundation-model inference.
  • Email and messaging delivery.
  • Payments and merchant services.
  • Industry-specific systems of record.

The integration boundary should remain typed, observable, retry-safe, and replaceable.

Self-hostable now

CPU-compatible web services, APIs, workers, dashboards, workflow engines, and open-source applications can be packaged as MIOSA deployments.

Applications that require persistent graphical sessions may run in MIOSA computers instead.

Confirm image architecture, memory, disk, network, health, and license requirements before promising support for a specific upstream project.

Coming soon

Future capabilities belong in product planning and architecture, not in a current-tense setup guide.

The build guides may explain how to prepare an application boundary for those capabilities.

They must not claim the capability shipped until CLI help, API documentation, acceptance coverage, and a production result exist.

How to read an example

When an example says you can build a video studio, it means:

  1. MIOSA can host the application, data, storage, workflow, customer isolation, review experience, and deployment.
  2. Video generation can use a connected provider today.
  3. Managed MIOSA GPU inference remains planned until this page marks it available.

That same distinction applies to voice media, private model serving, and other specialized runtimes.

Was this helpful?