Workspaces, encrypted BYOK, and an MCP server agents can drive.
Every project carries its own generation settings — text model, image model, voice, tone, vocabulary level, audience, target duration, CTA. Two channels with different personalities live side by side without leaking into each other.
A workspace is a hard boundary: its projects, keys, and channel connections are invisible to every other workspace. Invite teammates as editors (make reels) or admins (also manage settings, keys, and users).
Add your own Gemini, OpenAI, and ElevenLabs keys — encrypted at rest, used only for your generations, billed by the provider directly to you. No credits system, no markup, no lock-in.
The studio ships a built-in MCP server: connect Claude Desktop, Claude Code, or Cursor with a workspace-scoped token and your agent can plan, script, compose scenes, preview frames, render, and publish — the same guarded pipeline you use, under your review.
Renders queue fairly with visible position, and each workspace has a daily render budget — so an over-enthusiastic agent or script can never melt the studio or surprise anyone.
Images, voices, and renders keep their previous versions when regenerated — experiments are reversible, and nothing good is ever lost to a retry.
The MCP server isn't a demo bolted on the side — it's the same guarded pipeline the editor uses, exposed as tools. A workspace-scoped token gives your agent exactly your workspace: your projects, your keys, your review step. Nothing posts unattended; the agent drafts and composes, and you approve.
The guardrails are structural, not promises. Renders queue fairly and every workspace has a daily render budget, so a runaway script can't melt the studio or your bill. Keys are encrypted at rest and billed by the provider directly to you. And every generated asset keeps its previous versions — an agent experiment is always reversible.
Workspaces, encrypted BYOK, and the MCP server are in the free tier. No feature gates.
Start free