Civitai OrchestrationSubmit AI workflows. Get results.

Video, image, audio, text — one API, many providers. Build against the same orchestrator that powers Civitai.

Workflows, not endpoints

Describe the work you want done — the orchestrator picks a provider, routes the job, and streams results back. You don't manage capacity.

Multi-provider by default

FAL, Google, Bytedance, Civitai workers — the orchestrator races providers and selects the best fit for each job.

Typed recipe catalog

One recipe per job type (video-gen, image-gen, upscaling, transcription, TTS…) with validated inputs and predictable outputs.

Sync or async

Poll, subscribe, or wait inline with the `wait=` parameter. Webhooks supported for production integrations.

MCP-native

Connect Claude Desktop, claude.ai, or any MCP-aware client to the same orchestrator. Find a service, get its exact input schema and run it, over HTTP at /mcp/v2.

Where to start

Get started — authenticate and submit your first workflow.
Guide — workflows, submitting work, results and webhooks, errors and retries.
API reference — every consumer-facing operation, generated from the OpenAPI specification.
Recipes — one page per job type, with validated inputs and predictable outputs.
MCP server — connect Claude Desktop, claude.ai, or any MCP-aware client to the same orchestrator.

Guide and recipe pages are also served as plain markdown — append .md to the URL — and /llms.txt indexes them for agents. The generated per-operation reference pages under /orchestration/reference/operations/ are the exception: they are HTML only.