# How to connect cuttingfloor: AI apps, REST and MCP

> Edit video in the AI app you already use (Kazi, ChatGPT, Claude Cowork, Codex or your own), or call every editing tool from your code with an API key, over REST or MCP.

## I make videos (creators, students, marketers)

Use the AI app you already have. You sign in once: no keys, no code.

### Kazi

_Fastest start_

Our own AI agent, on your phone or computer. It opens with a sample clip and edits it on cuttingfloor. Nothing to set up.

[Try it live, free](https://app.apifunnel.ai/s/2XDlD9QfXx) · [How it works](https://cuttingfloor.ai/agents/kazi)

### ChatGPT

Add cuttingfloor to ChatGPT once, then ask for the edit in a chat: “cut the pauses out of this”.

[Set it up](https://cuttingfloor.ai/connect#chatgpt)

### Claude Cowork

Give Claude your footage and what you want made. It works through the edit as a task while you get on with other things.

[Set it up](https://cuttingfloor.ai/connect#claude-cowork)

### Codex

In the Codex app, on the command line or in your code editor: one line adds cuttingfloor, and you sign in once.

[Set it up](https://cuttingfloor.ai/connect#codex)

### Your own AI app

Claude, Cursor, VS Code, Goose or any app that takes an MCP server. Paste one address and the app opens the sign-in for you.

The address:

```
https://api.cuttingfloor.ai/mcp
```

[Pick your app](https://cuttingfloor.ai/connect#apps)

## I build with code (recommended for engineers and builders)

One API key from your account page, two ways to use it. It acts as your account and uses its credits.

### REST API

_Most efficient for LLMs using code execution_

Every editing tool is one HTTP call: POST https://api.cuttingfloor.ai/v1/<tool> with its arguments as JSON. Code an LLM writes calls the tools directly and chains them in one run, without loading every tool’s schema into its context. The OpenAPI spec feeds code generators and API clients.

A first call:

```
curl https://api.cuttingfloor.ai/v1/timeline.create \
  -H 'Authorization: Bearer cf_…' \
  -H 'Content-Type: application/json' \
  -d '{"name": "First edit", "sequence": {"width": 1080, "height": 1920, "fps": {"num": 30, "den": 1}}}'
```

[Make an API key](https://cuttingfloor.ai/account#api-keys) · [OpenAPI spec](https://api.cuttingfloor.ai/openapi.json)

### MCP with an API key

The full MCP server for your own agent, a back-end job or anything with no browser to sign in: send the key as a header. The same key works for REST, and both act as your account and use its credits.

Your app’s MCP settings:

```
{
  "mcpServers": {
    "cuttingfloor": {
      "url": "https://api.cuttingfloor.ai/mcp",
      "headers": {
        "Authorization": "Bearer cf_…"
      }
    }
  }
}
```

[Make an API key](https://cuttingfloor.ai/account#api-keys) · [Setup for each app](https://cuttingfloor.ai/connect#any)
