Skip to content

Emberu with AI agents

Let an AI assistant caption a timeline, move clips, place graphics and render the result.

What an assistant can do

Emberu gives an assistant its own operations: captions, the timeline, graphics and export. Edits land in the same undo history you use. Agent access comes with Pro and Max.

A first request to type. "Caption this timeline, use the brand Animation, add zooms and transitions, then export to this folder."

Read and inspect

Orientation
What works right now, what is blocking it, and which version of Emberu is running.
Projects and media
Which projects exist, what the media pool holds, and which timeline is open.
Review
Stills at a time or a spoken word, contact sheets over a span, and short previews with audio.

Edit and place

Captions
Generate captions, then edit text, timing, lanes, Animations, Sparks, Links and Formations in one batch.
Timeline
Trim, split, replace, retime, reframe or ripple-edit clips, and place or edit graphics.
Audio
Report the mix's sounds and levels, trim pauses, and place sound effects.
Colour
Grade a clip, match cameras, apply LUTs, and read scopes.

Run and deliver

App and settings
Open a page, change the Generate, Ignite and Export options, and install Ember Engine.
Delivery
Run Ignite, adjust or discard its result, and render clips from a Resolve timeline or an Emberu project.

With Resolve connected, the assistant gets the Resolve tools and works on the active timeline. With an Emberu project open, it gets the local editing tools and works there.

Connecting an assistant

  1. Open Settings > Account > Agents

    Emberu lists the assistants it can configure. Connect one.

  2. Or register from a terminal

    For an assistant Emberu does not list, run Ember mcp --register <assistant>. The same command takes --status and --unregister.

  3. Start a conversation

    Ask the assistant to use Emberu.

A row that reads Not installed has nothing to connect: install that assistant first.

Restart an assistant that was already open when you connected it.

How a good session runs

  1. Status first

    It checks Emberu's state before acting. Blockers name the fix, such as Resolve not running or no project open.

  2. Read before writing

    It reads the captions and clips it is about to change.

  3. One step per intent

    A group of changes lands as one step.

  4. Generate once

    Run Generate and Ignite once with the options you approved, then restyle instead of regenerating.

  5. Look at the frames

    It renders stills or a short preview, so composition problems show before the full render.

  6. Report the result

    The answer names what changed, any warnings, and where the files are.

Guardrails

  • It works on the project it opened and the files you point it at, and does not scan your disk.
  • A stale edit is refused instead of overwriting work that changed since it was read.
  • A step Emberu cannot verify is refused, never half-applied or swapped for a different result.
  • Changing page keeps a finished Ignite run; discarding it is a separate action.

Docs for agents

/llms.txt lists every topic with a one-line summary, so an assistant can find the right page before it starts a task. Inside Emberu it reads the app's own guide and tool fields, which are newer than this page.

Where the work happens

This documentation is indexed in plain text at /llms.txt.

Also worth reading