How to make videos with Claude or ChatGPT over MCP

MCP in plain words

MCP defines one common way for an AI application to connect to data, tools and workflows. A service publishes an MCP server with a set of tools, and a client such as Claude decides which tool to call to fulfil your request.

The standard is open and supported by Claude, ChatGPT, code editors and other clients. For Solistra it means you can direct an edit through an ordinary conversation with your own Claude or ChatGPT.

What Claude and ChatGPT can do in Motion

The assistant calls Solistra tools, and every action shows up right away on the preview and timeline.

• Create a project in 9:16, 1:1 or 16:9 with the length you need
• Use photos, videos, fonts, your logo and music from your library
• Build scenes with camera moves and transitions
• Add animated text and cut scenes on the beat of the music
• Generate new footage with Solistra image and video models
• Render a quick preview and a final MP4 in 1080p or 4K

Adding the connector

In Claude on the web or the desktop app, open connectors, choose to add a custom connector and paste the Solistra MCP server address from the Motion page. Then sign in to Solistra over OAuth and tick the permissions. The free Claude plan allows one custom connector, and on Team and Enterprise the organization owner adds it.

In Claude Code, add the server with "claude mcp add --transport http solistra" followed by the server address, then sign in with /mcp. Your Claude account is your own and paid separately. Claude is not available in every country, and Anthropic publishes the list of supported regions.

In ChatGPT, add a custom MCP server on chatgpt.com in the plugins section: the plus button, add a custom MCP server, the Solistra address and OAuth sign in, then create the plugin. In chat, call the plugin with @. An installed plugin also works in the ChatGPT apps, and in Business and Enterprise workspaces an admin allows it. ChatGPT asks you to confirm render and publishing actions first. ChatGPT is not available in every country either.

Permissions and spending control

Sign in uses OAuth 2.1 with PKCE, as the MCP specification requires for HTTP servers. On the connect screen you grant permissions with checkboxes: reading the library, creating scenes, rendering with credits. Publishing is off by default, and you set the daily spending limit.

The assistant receives short lived links to low resolution previews, not your source files. Solistra does not see your conversation with Claude or ChatGPT, only tool calls and short status updates. Requests through the connector go to Anthropic or OpenAI, so Solistra asks for consent to cross border transfer at sign in.

Example commands

Talk to Claude or ChatGPT the way you would brief an editor: format, length, mood, what to show at the start and at the end.

• "Make a 15 second 9:16 video for a coffee shop: three scenes from my library, cuts on the beat, logo at the end"
• "Generate a close up of a steaming cup and put it in as the second scene"
• "Make the text in the first scene larger and have it slide in from the bottom"
• "Pick a calm track from my library and align the cuts to the beats"
• "Show me the preview, and if it looks good, render it in 1080p"

Editing and undo

You can edit in words in the chat with Claude or ChatGPT or by hand on the timeline. Every action is saved as a separate step, and any step can be undone. The assistant reads notes you leave on scenes in the timeline at its next step.

How to do it in Solistra

Credits are spent only on generating footage and rendering, since the language model comes from your own Claude or ChatGPT. Credits are reserved before the final render.

• Open Motion and copy the MCP server address
• Add a custom connector in Claude or a custom MCP server in ChatGPT and sign in to Solistra
• Tick the permissions and set a daily spending limit
• Describe the video in chat, watch the preview and start the render