Make videos with Claude and AI agents
Motion Text has a hosted MCP server. Add one URL to your AI assistant, with no install and no API key, and it can draft complete lyric videos and story videos that open, fully editable, in this editor.
Most “AI video” tools give you a prompt box and a black box. Motion Text works differently. The agent drafts a real project, with timed captions, paced scenes, and a coherent visual treatment, then hands you a link. The link opens the full editor with the agent's work on the timeline, where you refine it, add your music and images, and export the MP4. Agents draft; you direct.
Add the connector
No install, no key. Pick your agent:
One command in your terminal:
claude mcp add --transport http motion-text https://motiontext.video/mcpAsk for a video
Give it words and a vibe: lyrics, a poem, a script with one beat per line. The agent replies with an edit link when the draft is ready.
Open the link and make it yours
The draft loads on a real timeline: retime anything, add music and images, apply presets, and export MP4. Drafts last 30 days. Save one to your free workspace to keep it for good.
Setup guide for your agent
Step-by-step connector instructions, config-file locations, and gotchas for each coding agent.
Claude
Paste one URL into Claude's connector settings and ask for a video in plain language.
Setup guideChatGPT
Add one connector URL and ChatGPT drafts editable videos. No plugin, no key.
Setup guideClaude Code
One command, and Claude Code drafts real, editable videos from the same terminal you code in.
Setup guideCodex
Add one entry to Codex's config and the OpenAI CLI drafts real, editable videos from your terminal.
Setup guideCursor
Add one MCP entry and Cursor's agent drafts real videos you finish in the Motion Text editor.
Setup guideVS Code
Add an HTTP MCP server and Copilot's agent mode drafts videos you finish in the editor.
Setup guideWindsurf
Point Windsurf's Cascade at one MCP URL and it drafts real, editable videos.
Setup guideCline
Add one server in Cline's MCP settings and the agent drafts real, editable videos.
Setup guideGuides
The workflow itself, rather than the connector: what to ask for, what comes back, and where you take over.
How to make a video with an AI agent
Give Claude, ChatGPT, Cursor, or any MCP client one URL, and it drafts a real project on a real timeline — not a black-box clip.
Read the guideHow to make a video from the terminal
One command in Claude Code, or two lines in Codex's config, and your CLI drafts videos you can render to MP4 without opening an editor.
Read the guideThe anonymous toolset
No token, no account: drafting only. Media, AI imagery, and rendering stay in the editor. The full toolset is a sign-in away, below.
| Tool | What it does |
|---|---|
create_story_draft | Build a complete lyric-video or story draft and get back an edit link. Give it a script and the Director paces it, or supply exact timings yourself |
list_visual_presets | The six complete looks: treatment, typography, colors, motion |
get_spec | Machine-readable capabilities, limits, and the draft schema |
Prompts that work well
Make a 30-second portrait lyric video called “Night Drive” from these lyrics, dark mood, fast pace: …
Turn this poem into a dreamy, slow quote video, one line per beat, and give me the edit link.
Create a 20-second story draft with three scenes and karaoke captions for this script: …
What visual presets does Motion Text have? Use the one that fits a horror teaser.
Sign in for the full studio
Add the studio endpoint and your client walks you through a one-click sign-in. There is no token to copy and no header to paste. Signed in, your agent moves from drafts to your actual workspace: sixteen tools covering projects, scenes, timed text (create, update, and delete), presets, Director cuts, and validation, plus real 1080p/4K server renders on Pro.
Add the studio endpoint the same way. When your agent prompts you to authenticate, sign in with Google and click Allow. Pick your agent:
One command in your terminal:
claude mcp add --transport http motion-text-studio https://motiontext.video/mcp/studioEverything the agent touches is scoped to your account: your projects, your plan's limits, watermark rules included. To disconnect, sign out in your agent or revoke access from the app's connector settings.
Questions
Which AI apps can connect?
Anything that speaks the Model Context Protocol over Streamable HTTP: Claude Code, OpenAI Codex, Cursor, VS Code (Copilot), Windsurf, Cline, Claude web and desktop, and most agent frameworks. The tabbed setup above covers each one. Anonymous access needs no key at all.
Do I need an account?
Not to create or open drafts. For the full toolset on your own projects, add the studio endpoint (motiontext.video/mcp/studio) and sign in when your client prompts. Signing in for the first time creates a free account. Nothing to copy or paste.
Can the agent render the MP4?
With an agent token on a Pro account, yes. submit_story_render queues a real 1080p or 4K server render, and get_story_render_job polls it to a download link. Anonymously, rendering stays in the editor: the agent drafts, you export.
Can the agent add music or images?
Not remotely. Media uploads and AI imagery happen in the editor, where you can see (and own) the result. Agents handle structure: timing, words, scenes, styling.
Can I edit what the agent makes?
That's the whole point. Drafts and projects live on a real timeline. Drag any caption, retime any scene, restyle everything, then export.
Is there an HTTP API for agents that don't speak MCP?
Yes: POST /api/drafts with the same JSON body as the create_story_draft tool returns a draft link, and GET /api/spec describes everything. See /llms.txt for the short version.
Agents draft. You direct.
Or skip the robots entirely. The editor is free in your browser.
Open the editor