How to connect ChatGPT to Vyra

Add Vyra's MCP server, authorize it, and ChatGPT can list your projects, read the footage, build a cut, add captions and music, and export. Same prompts as Vyra's own chat, plus you can ask for a plan first.

Before you start

• A Vyra account on an MCP plan (MCP Starter or MCP Pro).
• A ChatGPT plan that supports MCP connectors. The desktop path uses the Codex-compatible config.
• The Vyra MCP server URL is https://api.usevyra.com/mcp. Details at /docs/mcp.
• A project with footage already uploaded.

Steps

1. Open the Codex MCP config. ChatGPT's desktop path and other Codex-compatible clients read the same file.

2. Add the Vyra server and restart.

   [mcp_servers.vyra]
   url = "https://api.usevyra.com/mcp"

   Copy

3. Authorize. On first use the OAuth 2.1 flow opens in your browser. Sign in to Vyra and approve.

4. Open a project in the browser. Go to https://app.usevyra.com, sign in with the same account, open a project. The tab binds itself and a green indicator shows top-right. Keep at least one Vyra tab open.

5. Start a chat and make sure the Vyra tools are on.

6. Ask for a plan before an edit. Have it summarize the footage and propose a structure.

7. Edit in steps. Rough cut, captions, music, export. One message each.

8. Verify. Ask it to read the timeline back and capture a few frames.

The prompt to use

Using the Vyra connector, open my project "launch video" and summarize the footage. Propose a 30-second hook-and-demo structure. Me stating the problem, screen recording of the fix, me showing the result. Wait for my OK, then build it with phrase captions and a lo-fi track ducked under my voice.

Check the result

• The summary names real clips and lines from your project.
• The timeline read-back matches what ChatGPT said it did.

FAQ

ChatGPT said it made the edit but nothing changed. Ask it to read the timeline back. Then say "That's not what the timeline shows. Do X now."

Can I use ChatGPT and Claude on the same project? Yes. Same project either way.

Related

• /docs/mcp
• /chatgpt
• /for/chatgpt-users
• /prompts/troubleshooting