Agent control (MCP)
VidVertex ships an MCP server — the open standard AI assistants use to operate software. Connect it once (how) and an assistant like Claude can run the app the way you do: set up brands and pools, build stacks, write post texts, render, check the result frame by frame, and schedule. You describe the outcome; the assistant makes the tool calls.
Everything the assistant does lands in the same data folder the app uses. Open windows update live, and while an assistant works the app shows a slim strip along the top naming the client and its current action, with a click-open feed of everything it did. There is deliberately no "stop" button — the app cannot halt a client — and no unfinished actions are ever hidden.
01ConnectingClaude Desktop and Cowork install a one-click extension (.mcpb) exported from Settings → Integrations; Claude Code and Codex register one command; any other…02What the assistant may do — and what it may notThe same guardrails as every unattended context apply (see Automation)03A tool reports what it didEvery write tool answers with its effect, not just "ok": placing footage returns where it landed on the canvas and warns when the placement loses picture;…04The tools105 calls, mirroring the app surface.05What this looks like in practice"Import my Zernio profiles as brands, set them all to 18:00 Europe/Berlin, and point their background slots at the 'Gym B-roll' pool." — three calls, any…