Connect your MCP client

MCP gives your assistant tools for creating and editing Ghostmotion videos. The client discovers available tools and their input schemas from the server. Connect by signing in from your assistant, or with a workspace API key.

Sign in from ChatGPT, Claude or another assistant

Assistants that support MCP sign-in (OAuth) need no API key. Add the server URL and the assistant sends you to Ghostmotion to approve it.

https://app.ghostmotion.ai/mcp

• ChatGPT: in Settings → Apps & Connectors → Advanced, turn on Developer mode, then create a connector with this URL and OAuth authentication.
• Claude: in Settings → Connectors, add a custom connector with this URL.
• Claude Code, Cursor and other clients: add the URL without an Authorization header, then sign in when the client asks. In Claude Code, run /mcp and choose Ghostmotion.

Ghostmotion shows the app, your current team and what it may do. A reviewer can approve read-only access. Click Connect. You can also set a spending limit; without one it uses the team’s credits as you would, within each run’s cap. The connection then appears in Settings → Integrations, where you can see what it has used and disconnect it. In ChatGPT you can also attach files to a request; Ghostmotion uses them as references.

Or connect with an API key

For clients without sign-in, scripts and automation, use a workspace API key in the Authorization header.

1. Create a workspace API key

Choose read permission to list projects and get results. Add write permission to create or edit videos. Set a spending limit that fits this integration.

2. Configure the server

Transport

Streamable HTTP

https://app.ghostmotion.ai/mcp

Authorization: Bearer YOUR_API_KEY

Replace YOUR_API_KEY with your key in the client’s connection settings. For separate header fields, the name is Authorization and the value is Bearer YOUR_API_KEY. Keep the key out of chat messages.

Your client must support Streamable HTTP with a custom Authorization header. The exact settings format depends on the client.

For a ready-made command, use the Claude Code or Codex guide.

3. Check the connection

Refresh the client’s tool list. Ask it: “List my Ghostmotion projects.” A successful response, including an empty list, verifies read access without starting a paid run.

• create_videoCreate a video
• send_messageEdit or continue a video
• get_runCheck a run and get results
• list_projectsList projects
• list_charactersList characters
• create_characterCreate a character
• stop_runStop a run
• prepare_reference_uploadPrepare a reference upload
• complete_reference_uploadComplete a direct reference upload

4. Ask for a video and follow the run

Use Ghostmotion to create a 20-second launch video for our design app. Ask me for the product details first. After starting, check the run until it finishes and return the video link.

create_video returns a run ID immediately. The client should wait for poll_after_seconds, then call get_run until the run ends. Check reply and new_films; the creator can ask a question without publishing a new film.

Use send_message for edits or answers. See the complete tool reference for inputs, results and errors.

Characters

Create a recurring AI character (a photorealistic person who is not a real individual) or an illustrated character with create_character: a name, a kind and a description of their look and voice, plus reference images if you have them. It returns a run ID: poll get_run until character.ready is true. Then cast them in a video with cast: [{ "character_id": "…", "role": "lead presenter" }] on create_video or send_message. list_characters shows the team’s characters and which are ready. Digital twins of real people are created in Ghostmotion, where the person gives their consent; once made, they can be cast here too.