Start free

AI agents

Agentic Video Editing

Describe the video and an agent makes it: Director in Studio, or your own agent through the MCP server or the CLI and skill. Check it in Studio, then render it once or for every record in your data.

Start free Connect your agent
  • Ikea
  • McDonald's
  • Twitter
  • Randstad
  • Nike
  • Coca-Cola
  • Zillow
  • Angel
  • ClickFunnels
  • TheSoul Publishing

Your agent writes the video. Shotstack previews, renders and reuses it.

An AI agent can plan a video and write every clip, caption and transition. What it can't do on its own is preview the result, let a person fix it, or produce the same design again tomorrow with different data.

Shotstack gives it all three. The agent writes an edit, the same JSON our render API, Studio editor and templates use. That edit opens in the browser for review, renders in the cloud, and stays editable as a template you can run again.

Your agent

Use Director, or connect your agent through the MCP server or CLI

  • No setup

    Director

    Describe it, Director makes it. Tell Director, our AI editor, what you want, and it builds the video in Studio, right in your browser.

    Open Studio

  • Coding agents

    CLI and skill

    Your coding agent sets up Shotstack and makes your first video. One paste installs the CLI and its skill, then hands Claude Code, Cursor, Codex or Grok CLI the task.

    CLI guide

  • AI assistants

    MCP server

    Your AI assistant makes videos when you ask. Add the MCP server to Claude, ChatGPT or Cursor, and in chat it shows Studio right in the conversation.

    MCP server guide

Connect your agent

How it works

From a prompt to a video you can reuse

  • It writes the edit

    Describe the video. The agent writes the edit as JSON, the format our API and Studio share, and checks it for errors before anything renders.

    How an edit works

  • You check it in Studio

    Open the draft in the browser editor, play it and change anything by hand. Previewing costs nothing; credits are only used when you render.

  • It renders once, or for every row

    Render an MP4, or save the edit as a template with merge fields and render it for every product, listing or customer, with no model in the loop. Or let the agent write a new edit for every video.

What your agent can work with

  • Generate

    AI video, images, voice and music

    Give an image, video or audio asset a prompt and a model, and Shotstack generates it while the video renders. An unchanged prompt reuses its asset at no charge.

  • Bring

    Your own media

    Give the agent your images, footage and audio by URL or upload, and it places them on the timeline.

  • Compose

    Text, captions and motion

    Rich text in built-in or your own fonts, captions transcribed from the voice-over, SVG and HTML5 graphics, and keyframed position, scale, rotation and opacity.

Made by an agent

Built by an agent from a short brief

Each of these was built by an AI agent from a short brief and rendered by Shotstack. Open one in Studio and make it yours.

  • Image and video generation

    Product ad from one photo

    Generated close-ups of a bag from a single product photo, its details called out, then the price.

    Open in Studio
  • Keynote launch

    Product reveal

    A launch in the style of a keynote: the product's name, three lines on what it does, and its price.

    Open in Studio
  • Sports broadcast

    Full-time match recap

    The score, the match stats and how the goals went in, laid out like a broadcast from one match's data.

    Open in Studio

Motion design showcase

Trusted to deliver

Join companies large and small rendering thousands of videos every day

Over 100,000 Businesses from 119 Countries Trust Shotstack

I used to edit the JSON myself. Now I describe what I want, and an agent sets it up.
Michael Peloquin Borderline Media

Questions

What is agentic video editing?
An AI agent turns a request into a finished video. It plans the video, writes the edit (the clips, text, timing and effects) as JSON, and calls tools to preview and render it. Shotstack is the tool it calls: it validates the edit, shows it in a browser editor and renders it in the cloud.
Which AI agents work with Shotstack?
Director, the AI editor built into Studio, needs no setup. Any client that supports MCP can use the Shotstack MCP server, including Claude, ChatGPT, Cursor, VS Code Copilot and Gemini CLI. Coding agents can use the Shotstack CLI and its skill. You can also give your own agent a render tool that calls the Edit API directly.
Do I need an API key to try it?
Not to start. With the CLI, checking an edit and opening it in Studio need no key and no credits. Connecting the MCP server needs a sign-in. Rendering needs an API key: renders in the stage environment are free and watermarked, and production renders use credits.
Can a person edit what the agent made?
Yes. The agent's edit opens in Studio, where anyone can play it, move clips and change text before rendering. Saved as a template, it can be opened and changed again later.
Will the same edit produce the same video?
A saved edit or template renders the same layout and timing every time, with only the merge field values changing. An AI asset with an unchanged prompt, model and options reuses the asset it generated before.
What output formats and resolutions does Shotstack support?
Shotstack renders video to MP4 or GIF, audio to MP3, and stills to JPG, PNG or BMP. Choose a resolution preset from preview up to 1080 and 4K, or set custom dimensions for vertical and square video.
Does Shotstack have an MCP server?
Yes. The Shotstack MCP server runs at mcp.shotstack.io over Streamable HTTP and signs in with OAuth or an API key header. It gives any MCP client the tools to write, check and render an edit, generate AI assets and open the result in Studio. Setup is one command or one URL in the client settings.
Can Claude make videos?
Claude has no video model of its own, but it can make videos through a tool. Connect the Shotstack MCP server to Claude.ai, Claude Desktop or Claude Code, and Claude writes the edit, opens it in Studio for you to check and renders it to an MP4. Claude Code can also use the Shotstack CLI and its skill.
How do I build my own agent that makes videos?
Give your agent a render tool that sends a JSON edit to the Edit API and a status tool that polls for the finished video URL. How to build an AI video agent walks through the tool schema, the agent loop and the render function in Python. If you would rather not write the loop, the MCP server and the CLI give an agent you already use the same tools.

Guides

Your first render in minutes

  • Free tier: 10 credits
  • No credit card required
Start free Connect your agent