Workflow automation

Oraphim automation is built around explicit MCP tools and workflows, not an invisible script that can silently rewrite the project. A reliable automation reads the current state, performs bounded operations, validates the result, and leaves evidence that a human can inspect.

Good automation candidates

Use automation for repetitive structural tasks: creating or renaming project objects, building known graph structures, applying consistent metadata, checking project state, producing batches of export jobs, or running a repeatable authoring workflow across well-defined selections.

Avoid turning an ambiguous creative decision into a large unattended mutation. For example, "make this whole film better" is not a good transactional instruction; "create a 1080p review export for these three named timelines" is bounded and verifiable.

Safe pattern

1. Query project/capability state first.
2. Confirm the intended resource IDs or names.
3. Save a project milestone.
4. Execute the smallest workflow that accomplishes the task.
5. Re-query or validate the changed objects.
6. Inspect output in Oraphim or through a captured/rendered artifact.
7. Save only after the result is accepted.

Example: title plus review export

An MCP client can create a composition, add/edit a text layer, place or bind that composition where required, save the project, then add a review export to the Render Queue. The important part is not the prompt wording; it is that every stage maps to a real Oraphim resource and can be inspected after the call.

The installer/repository includes connection examples and a create_title_and_export.jsonl example for integration testing.

Headless/service operation

Oraphim also supports a service-mode MCP runtime used for certified headless authoring and testing. Service mode is intended for integrations that know exactly which project/tool operations they need. It should not be treated as permission to skip normal project validation or final render checks.

Verification

For automated exports, independently inspect the final media. For authored graphs/timelines, save and reopen the .orp and validate the expected resource identities and connections. Automation is complete only when the persisted project and delivered artifact both match the request.