MCP connector
Connect Claude, ChatGPT, Codex or Cursor to VideoDraft and make media without leaving the chat.
On this page
What the connector does#
The connector lets an AI app you already use work with your VideoDraft account. From the chat it can make images, video, voice, music and 3D models, run a video project, and read your projects and media.
It uses the same account, credits and history as the web app. Images, video and sound it makes appear in AI Studio. It runs on VideoDraft servers, so there is nothing to install. The CLI is the terminal version of the same tools.
| Setting | Value |
|---|---|
| Connector name | VideoDraft |
| Server URL | https://app.videodraft.ai/api/mcp |
| Transport | Streamable HTTP |
| Sign-in | OAuth in your browser, or a personal access token |
| Plans | Every plan, Starter included |
Claude and Claude Code#
- 1In the Claude app, open Customize, then Connectors, then Add custom connector.
- 2Enter the name VideoDraft and the server URL above.
- 3Click Connect and finish signing in to VideoDraft in your browser.
In Claude Code, add the server from your terminal, then run /mcp in a session to sign in.
claude mcp add --transport http videodraft https://app.videodraft.ai/api/mcpCodex#
Add the server with this command, or in Codex’s MCP settings. Sign in when Codex asks.
codex mcp add VideoDraft --url https://app.videodraft.ai/api/mcpChatGPT and Cursor#
ChatGPT: turn on Developer mode, then add a custom app with the server URL and OAuth. Your ChatGPT plan and workspace must allow custom apps. OpenAI’s guide, linked at the end of this page, has the current steps.
Cursor: open the command palette, run Open MCP settings and add the entry below. Restart Cursor and sign in.
{
"mcpServers": {
"videodraft": { "url": "https://app.videodraft.ai/api/mcp" }
}
}Use a token instead of a browser sign-in#
If an app cannot sign in through the browser, create a personal access token at app.videodraft.ai/mcp-keys and send it as a bearer header. Tokens start with vd_mcp_.
claude mcp add --transport http videodraft \
https://app.videodraft.ai/api/mcp \
--header "Authorization: Bearer vd_mcp_…"You see a token once
VideoDraft keeps only a fingerprint of it, so it cannot show it again. If you lose it, create a new one.
What your agent can reach#
Tools come in groups called lanes. Agents call get_tool_catalog with a lane name to get its tools and how to use them.
| Lane | Covers |
|---|---|
| assets | Images, video, voice, music, sound effects, dialogue, 3D |
| asset_io | Uploads, downloads and media descriptions |
| projects | Create, read, update and version projects |
| project_data | Scenes, shots, styles, visual assets, timeline clips |
| production | Hosted production runs and video export |
| library | Media, generation history, characters, products, styles |
| account | Identity, credit balance, model lists, per-model costs |
Credits and access#
- Generations spend the same credits as the web app and the Mac app. The connector itself costs nothing.
- Ask for an estimate first. Video, batches of shot images, upscales and avatar renders cost the most.
- Media made outside a video project goes into one AI Studio project for that connection.
- Every call runs as you and can reach only your projects.
- A sign-in lasts 30 days and renews in apps that support it. Revoke a sign-in or token at app.videodraft.ai/mcp-keys, where sign-ins appear as OAuth: <app name>.
- If a request is rate limited, wait and try again. Do not resubmit the same job over and over.
Detailed editing happens in VideoDraft Editor
The connector can make media, produce a video and place clips on a web project timeline. Precise cuts, color, captions and mixing happen in VideoDraft Editor on the Mac.
Try it#
Estimate one image of a blue ceramic cup on an oak desk. After I approve, make it, wait until it finishes, and tell me the AI Studio project and the image link.Open that project in AI Studio to see and download the image. If the chat disconnects while the job runs, ask the agent to check the same job instead of starting a new one.