In beta · no waitlist

Video composition API, built for AI agents.

Your agent connects over MCP, REST, or the CLI — ingests footage, composes the edit as data, then renders production-quality video. You direct; it does the legwork.

Paste into Claude Code, Cursor, or Codex
set up https://cueframe.ai/skill.md
Start freeSee how it works

25 free renders a month · no card · no API keys — or run npx -y cueframe install

composition · data in
render · video out
{
  clips: [{ media: "med_x9", range: [12.4, 21.8] }],
  reframe: { follow: "active-speaker", aspect: "9:16" },
  overlays: [{
    primitive: "kinetic-type", text: "Ship it",
    zPlane: "behind-subject"
  }],
  captions: { style: "word-pop", brand: "bk_studio" }
}
Capabilities

Content-aware by default.

The motion-graphics craft an editor does by hand — reframing, captions, brand, graphics — exposed as parameters your agent sets, then scored by a judge.

Frames itself around the speaker

Face and active-speaker tracking reframe the shot across the whole cut — a moving camera, not a static crop.

reframe

Text behind the subject

Titles and captions rendered behind the speaker — the subject matte segmented at render, no manual masking. Content-aware depth, resolved from the footage itself.

zPlane: behind-subject

One compose, every aspect ratio

9:16, 16:9, 1:1, 4:5 — baked from a single composition, never re-edited per platform.

format

Captions & motion graphics

Word-level captions, lower-thirds, titles — composited at render as parameters, not hand-keyed.

setCaptionStyle

Custom graphics, in code

Author a TSX component, preview it, render it as an overlay — even behind the subject. Compile errors come back inline.

create_component
brand kit

Brand, bound at render

Fonts, colors, and logo resolve from a brand kit at render — every clip on-brand, zero re-authoring.

brandKitId
The loop

Three steps, raw media to render.

The same loop a human editor drives — every step an agent can run on its own.

1

Ingest

Pull footage from any public URL, or generate it from a prompt. CueFrame transcribes and detects faces so the edit lines up with the content.

$ cueframe media import https://… -p pro_x9
2

Compose

Author the edit as data — or hand over the intent and CueFrame drafts several cuts, scores each one, and keeps the winner. Validation is a free dry-run, before you spend a cent on a render.

$ cueframe compose pro_x9 --from-composition
3

Render

Render production-quality video to any aspect ratio, stream the progress, and export to Premiere or Final Cut.

$ cueframe render pro_x9 -o short.mp4
Interfaces

Reach it however your agent runs.

One composition model behind every door — the same typed operations over MCP, HTTP, or the terminal.

#01 · MCP

Hosted MCP

Point your agent at the MCP endpoint. Zero install — it discovers every tool, typed and fail-loud.

claude mcp add --transport http cueframe https://api.cueframe.ai/v1/mcp
#02 · REST

Live OpenAPI

A discoverable OpenAPI surface. Read the contract, validate against it, render — no SDK guesswork.

curl https://api.cueframe.ai/v1/openapi.json
#03 · CLI

Command line

Drive it from the terminal or any script. The same operations, piped into any workflow.

npm i -g cueframe

Full setup — endpoints, auth, and API keys — in the setup guide.

By the numbers
4
Ratios, one compose
9:16 · 16:9 · 1:1 · 4:5
4K
Render output
H.264 MP4, up to 2160p
$0
Dry-run validation
fails loud before you spend
100%
Reproducible
same inputs, same render
Pricing

Usage-based, no surprises.

Start free. Paid tiers include a generous render bucket, then simple pay-as-you-go overage — a render has real cost, so we price it honestly instead of pretending it's unlimited. Annual saves you two months.

Free

Build and ship your first agent video.
Free
  • 25 video exports/mo
  • 10 GB cloud storage
  • 200K AI tokens/mo
  • 60 transcription minutes/mo
  • 5 AI Director runs/mo
  • Brand kits
Start free
Most popular

Pro

For creators and solo builders shipping regularly.
$50/mo

or $480/yr annually

  • 400 video exports/mo, then $0.40 each
  • 50 GB cloud storage
  • 2M AI tokens/mo
  • 1,000 transcription minutes/mo
  • 200 AI Director runs/mo
  • AI generation — $2.00/generation
  • Brand kits
Choose Pro

Business

For teams producing at volume.
$100/mo

or $960/yr annually

  • 900 video exports/mo, then $0.35 each
  • 200 GB cloud storage
  • 5M AI tokens/mo
  • 3,000 transcription minutes/mo
  • 500 AI Director runs/mo
  • AI generation — $2.00/generation
  • Brand kits
  • Video understanding
  • Priority processing
Choose Business

Studio & Enterprise

For high-throughput pipelines. Studio from $200/mo (or $1,920/yr annually). Enterprise — Custom volume, SSO, SLA, and dedicated support.

Choose StudioContact us
FAQ

Questions, answered.

What exactly is CueFrame?

A video-composition API. Your agent authors the edit as structured data — clips, reframe, captions, brand — then validates and renders it to production-quality video. CueFrame is the reliable hands; your agent brings the intent.

How does an agent drive it?

Three ways, same typed operations: a hosted MCP endpoint (zero install), a live OpenAPI surface over HTTP, or the CLI. The agent discovers the contract, validates against it, and renders.

Is it free?

Free to start — create a free account, no card required. Paid plans (see Pricing) unlock higher limits. Agents can also pay per call with x402, the machine-to-machine payment protocol: the agent signs each request from its own wallet, no account needed.

What can it output?

Vertical, square, or wide — up to 4K H.264 MP4. Or eject the timeline to Premiere or Final Cut. One composition, every aspect ratio.

Do you validate before charging for a render?

Yes. validate is a free dry-run against the live schema — it fails loud on bad input before a cent is spent on a render.

In beta

Give your agents a cutting room.

Paste one line into your agent and it wires itself up — no keys, no waitlist. Or start free in the console.

set up https://cueframe.ai/skill.md
Start freeRead the setup guide

Want to talk first? Book a 30-min call·Browse the API