> ## Documentation Index
> Fetch the complete documentation index at: https://www.hedra.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Hedra infrastructure and availability

> Understand Hedra's managed inference API, World Model Runtime, model adaptation, and deployment options, including research-preview access.

Hedra provides infrastructure for running and adapting multimodal models.
Developers can start with managed inference through the v3 API, using a shared
API wallet and model-specific contracts. The broader platform connects inference,
model adaptation, and compute for custom workloads.

## What can I integrate with today?

| Workload | Integration | Getting started |
| - | - | - |
| Image understanding and text reasoning | Supported models through `/v3/chat/completions`; compatibility routes also include `/v3/responses` and `/v3/anthropic/v1/messages` | [Vision and reasoning quickstart](https://www.hedra.com/docs/pages/developer/v3/quickstart.md) |
| Image, video, and audio generation | Asynchronous jobs through `/v3/models/{model}` | [Model-job quickstart](https://www.hedra.com/docs/pages/developer/v3/model-jobs.md) |
| Model adaptation and evaluation | Hedra Training, a research preview | [Discuss access](https://www.hedra.com/enterprise) |
| Isolated compute for custom code and pipelines | GPU Sandboxes, a research preview | [Discuss your workload](https://www.hedra.com/enterprise) |

Use the [model catalog](https://www.hedra.com/docs/pages/developer/v3/model-catalog.md) to confirm a model's
current inputs, outputs, limits, and price. Availability of a catalog model does
not imply that every modality or feature is supported through every endpoint.

## What is World Model Runtime?

World Model Runtime is Hedra's engine for multimodal workloads. It is part of the
infrastructure behind the platform. Developers using managed inference integrate
with the documented API routes and model contracts.

See [World Model Runtime](https://www.hedra.com/wmr) for the runtime overview.
For a workload involving a world-action or vision-language-action model, discuss
the model and serving requirements with the engineering team. The image-understanding
quickstart returns text; it is not a robot-control or world-action endpoint.

## How do deployment options differ?

Managed catalog inference is the self-service API path described in these docs.
Bringing your own model or deploying on your own infrastructure has a separate
onboarding process. See [deployment options](https://www.hedra.com/inference)
and [contact engineering](https://www.hedra.com/enterprise) with your model,
expected input/output sizes, throughput, latency target, and infrastructure requirements.

## How does this relate to Studio and the Creative Agent?

Studio is Hedra's creative workspace. Its help pages remain under **Resources →
Studio help**. The developer platform uses API keys and a prepaid USD API wallet;
Studio subscription credits do not fund v3 API requests. See
[API billing](https://www.hedra.com/docs/pages/developer/v3/billing.md).

Connecting an assistant to Hedra's MCP server is also different from running the
assistant's language model on Hedra. See [MCP setup](https://www.hedra.com/docs/pages/developer/v3/mcp.md) for
connected tools and [Coding agents](https://www.hedra.com/docs/pages/developer/v3/coding-agents.md) for the
language-model provider configuration.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.