Edit videos with Codex (ChatGPT)

The official ChatGPT desktop app supports MCP tools (Codex). Drop in PandaStudio's MCP config once and you can drive the editor from your ChatGPT conversation, on any plan that includes Codex.

1

Install PandaStudio + the ChatGPT desktop app

PandaStudio from writepanda.ai. ChatGPT desktop from openai.com/chatgpt/download. Sign in with a paid ChatGPT account (Plus, Pro, Team, Enterprise) — Codex tool use requires a paid plan.

2

Install Node.js if missing

bash
node --version

Need 18+. Get it from nodejs.org — ChatGPT's Codex runtime spawns the PandaStudio MCP server via npx, which needs Node on your PATH.

3

Enable PandaStudio's automation server + copy the token

PandaStudio's automation server is localhost-only and gated by a bearer token. To find your token:

  1. Open PandaStudio on your machine.
  2. Go to Settings → Automation.
  3. Toggle Enable automation server on. A bearer token will appear — copy it.

The token is a random 32-byte secret stored in your OS keychain. It never leaves your machine — agents only need it because the local HTTP server requires it on every request.

4

Open ChatGPT's MCP config

In ChatGPT desktop, open Settings → Connectors → Custom MCP and click Edit config. ChatGPT writes the config to a JSON file:

macOS
~/Library/Application Support/ChatGPT/mcp-config.json
Windows
%APPDATA%\ChatGPT\mcp-config.json

If the menu wording differs in your build, search the ChatGPT settings page for "MCP" — the desktop app's Codex tool registry sits behind that label across every recent version.

5

Install the PandaStudio Skill

The Skill teaches ChatGPT PandaStudio's full command surface (transcript editing, captions, motion graphics, exports). Run once in your terminal:

bash
npx skills add kamskans/pandastudio-skills

The installer detects the ChatGPT desktop app and copies the Skill into its knowledge directory.

6

Paste in PandaStudio's MCP config

mcp-config.json
{
  "mcpServers": {
    "pandastudio": {
      "command": "npx",
      "args": ["-y", "@writepanda/mcp"],
      "env": {
        "PANDASTUDIO_TOKEN": "paste-your-token-here"
      }
    }
  }
}

Replace paste-your-token-here with the token from Step 3. Save the file and restart ChatGPT — the next time you open a chat, the PandaStudio tools appear in the Codex tool picker (the wrench icon next to the input).

7

Drive PandaStudio from a chat

Open a new ChatGPT conversation, click the tool picker, enable PandaStudio, and try:

  • "Open my latest recording, transcribe it, remove filler words and silences, then export an MP4."
  • "Create a new 9:16 project, import the file at ~/Recordings/launch.mov, add bold captions, and preview it."
  • "List all my projects from this week and tell me which one is longest."

ChatGPT will ask permission for each tool call the first time. Approve "always" for read-only verbs (system_status, project_list, transcript_get) to keep the flow snappy. Keep destructive verbs (project_delete, export_publish_youtube) on per-call approval.

Troubleshooting

  • PandaStudio doesn't appear in the tool picker — force-quit and re-open ChatGPT (Codex only reads the config on cold start). Then check Settings → Connectors → Custom MCP for an error icon next to PandaStudio.
  • "npx not found" in the error log — Node.js isn't on PATH for GUI-launched apps. On macOS, install Node via the official installer (not via a shell-only manager like nvm) so the PATH stays available outside Terminal.
  • "Unauthorized" on every tool call — the token from Step 3 was regenerated since you pasted. Open PandaStudio → Settings → Automation → Regenerate token and update the config.