Quickstart

Three commands. The skill teaches your agent the workflow; the CLI does the work; the account holds the credits.

$ npx skills add vokox-ai/vokox
$ npm i -g https://vokox.ai/cli/vokox-cli.tgz
$ vokox auth login

Then, in your agent: “Make a 15-second vertical ad for this product photo, English voice-over.” The agent drafts a plan, shows the cost and waits for your yes.

plan.json

One file per deliverable. Steps reference each other with @id; local files with file:. Independent steps run in parallel; a rerun skips finished steps.

{
  "name": "coffee-ad", "budget": { "maxCredits": 300 },
  "defaults": { "aspectRatio": "9:16" },
  "steps": [
    { "id": "hero",  "type": "image", "model": "auto:image.hq", "refs": ["file:./product.png"], "prompt": "…" },
    { "id": "clip1", "type": "video", "model": "auto:video.hq", "refs": ["@hero"], "duration": 5, "prompt": "slow dolly in; …" },
    { "id": "voice", "type": "tts",   "language": "en", "voice": "en-female-1", "text": "…" },
    { "id": "music", "type": "music", "prompt": "warm lo-fi, 90 bpm, no vocals", "duration": 30 },
    { "id": "final", "type": "compose", "timeline": {
      "clips": [{ "asset": "@clip1" }], "voice": { "asset": "@voice" },
      "music": { "asset": "@music", "volume": 0.2, "duck": true },
      "captions": { "from": "@voice", "style": "bold-bottom" } } }
  ]
}
$ vokox run plan.json --estimate
$ vokox run plan.json --yes --max-credits 300

Make one like this

Give a link to a Reel, TikTok, Short, Pin or X post (or a video file) up to 60 seconds. VokoX cuts it into shots, reads the frames and the speech, explains why it works and writes a plan that rebuilds the idea with your files and brief. The reference is deleted after the study. 10 credits for the study; the plan is priced like any other.

$ vokox like https://www.tiktok.com/@…/video/… --product ./can.png --brief "my iced tea, summer launch"
$ vokox run vokox-out/like.plan.json --estimate

In the browser: account → Make one like this. Over HTTP: POST /v1/jobs { "type": "analyze", "input": { "reference": "<link or asset id>", "refs": [<your files>], "brief": "…" } }, then { "type": "plan", "input": { "plan": <output.plan>, "refs": [...] } }.

MCP server

For claude.ai, ChatGPT, Cursor or Claude Code without the CLI. Same tools, same credits.

$ claude mcp add --transport http vokox https://vokox.ai/mcp --header "Authorization: Bearer sk_…"

Tools: list_models, estimate_cost, generate, compose, list_gallery, recreate, analyze_reference, run_plan, get_job, list_recent, get_balance, library_index, library_search, library_get. Create an API key in your account.

HTTP API

Bearer auth with an API key. Jobs are asynchronous: create, poll, download.

POST /v1/price          { "type": "video", "input": { "prompt": "…", "model": "auto:video.hq", "duration": 5 } }
POST /v1/jobs           same body → { id, status, creditsHeld }
GET  /v1/jobs/:id       → { status: queued|running|succeeded|failed, output: { assets: [{ id, url }] } }
POST /v1/uploads        multipart "file" → { id }   (use the id in "refs")
GET  /v1/me             → { balance, topupUrl }

Terms

You own the outputs you generate, subject to the licenses of the underlying models. Refunds, acceptable use and everything else are in the Terms of Service.

Privacy

We store your email, your prompts, the files you upload and the files we generate for you, so you can find them again. Prompts and reference files are sent only to the model provider that runs the job. Details, including how to delete your data, are in the Privacy Policy.