Generate videos with OpenAI Codex
Add Motionvid with codex mcp add, sign in once, and Codex can make and deliver videos as part of any task.
Set it up
~/.codex/config.toml
[mcp_servers.motionvid] url = "https://motionvid.ai/mcp"
- 1Run: codex mcp add motionvid --url https://motionvid.ai/mcp
- 2Sign in: codex mcp login motionvid
- 3Ask Codex for a video.
Headless: add bearer_token_env_var = "MOTIONVID_API_KEY" under the url and export that variable. Codex takes the name of the env var, not the key itself.
Example prompts
- “Make a 30 second explainer of our API for the docs homepage and save it to ./public.”
- “Make a 30 second explainer on how compound interest works, with voiceover and music.”
- “Turn https://yourapp.com into a 20 second vertical launch ad that ends on our logo.”
- “Make a 15 second animated bar chart of these numbers: 2023 1.2M, 2024 2.9M, 2025 6.4M.”
Tips
- Install the skill (npx skills add tolypash/motionvid-agent) so Codex follows the same steps every time.
- Headless: set bearer_token_env_var = "MOTIONVID_API_KEY" in ~/.codex/config.toml and export the key.
- Codex takes the name of the env var, not the key itself.
- Ask for the preview frames before the render to catch wording changes early.
Questions and troubleshooting
- Can Codex make videos?
- Yes, with the Motionvid MCP server Codex can generate a video from a prompt, check its progress and download the finished MP4.
- It says I need to sign in, or the tools are missing
- The server is added but not signed in. Run the sign-in step again and log in with your Motionvid account. The connector needs a paid plan; on a free account the tools answer with an upgrade message.
- The video is taking a while
- That's normal. A build writes the script, animates each scene and records the voiceover, usually a few minutes. Ask the agent to check the status again rather than starting a new video.