The next ideaMade for momentumStart building
15 sec9:16Product launch
“Create a cinematic product reveal with a confident hook, three benefits, and a direct CTA.”
Video infrastructure for the agent era
Your agent can now ship video.
Give Claude, Codex, Cursor, or your own agent a real video engine for composing, rendering, previewing, and delivering production-ready videos.
Built forAI agent buildersDevelopersAutomation teams
user → agent
Create a cinematic 9:16 launch video for our new AI workspace. Use three scenes, captions, and a confident CTA.
One prompt → one render pipeline
Open-source MCP Official registry Node SDK n8n + Make + Zapier
14MCP tools
20agent skills
250+templates
3render backends
1composition contract
Made with the engine
From one instruction to a finished video.
These are the kinds of outputs your agent can create when it has composition, rendering, and review tools in the same loop.
Work less manuallyShip more consistentlySee the difference
24 sec16:9SaaS explainer
“Turn these product benefits into a clean explainer with kinetic type and a soft blue palette.”
Stop the scrollMake it memorableTry it today
12 sec9:16Social story
“Build a fast social story with a bold first frame, captions, and a loopable ending.”
One templateEvery audienceOne render pipeline
50 variants1:1Batch campaign
“Generate localized product videos for every row in this campaign spreadsheet.”
What agents can do
The video layer between intent and output.
MCP tools handle actions. Skills add context. VideoJSON keeps every render deterministic.
01
Brief to storyboard
Turn an idea into a hook, story, proof, and CTA before a frame is rendered.
intent → BRIEF.md02
Compose with intent
Build scenes, layers, transitions, captions, and effects through a typed contract.
VideoJSON → scenes[]03
Render at scale
Queue one video or thousands of variants with status, credits, and signed delivery URLs.
job → mp4 + webhook04
Review and improve
Preview, capture, measure, and analyze layouts so agents can debug their own output.
render → inspect → fixThe agent-to-video loop
One contract. Every stage of the render.
Your agent does not need to understand a dozen disconnected APIs. It speaks one composition language from the first idea to the final MP4.
01
step 01
Describe
Your agent turns a simple brief into a structured video plan.
Create a 15s product launch video02
step 02
Compose
VideoJSON gives every scene, layer, effect, and timing a contract.
scenes: [hook, reveal, cta]03
step 03
Render
The same composition moves through the browser, server, or edge renderer.
job_01HQ... · 68%04
step 04
Ship
Poll the job, fetch a signed URL, and deliver the final MP4 to your workflow.
https://cdn.../video.mp4ASK YOUR AGENT
terminal
Create a 15-second vertical product launch video.
Use the product-promo template.
Add captions and a confident CTA.GET A STRUCTURED JOB
terminal
{
"jobId": "render_01HQ7K",
"status": "queued",
"format": "1080x1920",
"webhook": "ready"
}Predictable render economics
Know the credit cost before you render.
The agent loop stays reviewable when output size and render cost are explicit before a job starts.
The baseline is 1080p30 rendered video. Cost scales with duration, resolution, and frame rate.
Preview the render-cost estimate before an agent spends credits.15s · 1080p30≈ 0.25 credits
2 min · 4K60≈ 16 credits
Use the same contract for a single preview or a queue of variants.
Built as infrastructure
Not a prompt wrapper. A video engine.
Use the interface that matches your product: MCP for agents, SDK and REST for code, CLI for CI, and workflow nodes for automation.
Deterministic VideoJSON compositions Browser, server, and edge rendering Signed outputs and idempotent jobs
Read the architecture AI agentsClaude · Codex · Cursor
MCP · SDK · CLIOne developer interface
VideoJSONDeterministic composition
Render pipelineBrowser · Server · Edge
MP4 deliveryR2 · signed URL · webhook
Choose your path
Start where your workflow starts.
Same rendering engine. Different entry point.
Give your agent a video memory.
MCP tools handle actions. Deep skills teach your agent how to plan scenes, choose effects, and review its own output.
- MCP tools with typed inputs
- 20 domain skills
- Preview and debug loops
Video infrastructure without the editor UI.
Build repeatable rendering into your product with the SDK, REST API, templates, webhooks, and a deterministic composition contract.
- VideoJSON composition format
- Node SDK and CLI
- Idempotent render jobs
Turn rows, events, and prompts into videos.
Connect a trigger to a template, render variations at scale, and deliver the finished file where your team already works.
- n8n, Make, and Zapier
- Batch-ready templates
- Webhooks and signed URLs
The intelligence layer
Tools give agents hands. Skills give them taste.
Install the domain knowledge that helps an agent choose a structure, respect timing, pick compatible effects, and review its own work.
Router intent interview and workflow selectionDomain VideoJSON, animation, effects, templatesWorkflow promo, social, explainer, kinetic typeSub-agents parallel scene workers and review loops
Install agent skills One engine, every interface
Connect the way your team already works.
Use the direct agent interface or connect your existing automation stack.
MCP Servermcp
Node.js SDKsdk
CLIcli
Zapierworkflow
n8nworkflow
Makeworkflow
Start with one render
Tell your agent what to make.
Install once. Give it a brief. Get a video back.