Use WideCast with Claude, ChatGPT, or Gemini

Turn "make me a video about X" into a real, finished video — by talking to your AI assistant in plain language. You connect WideCast once, optionally add our writing Skills, and then the AI writes a great script and produces the video. No coding.

What you'll need (5 minutes): a WideCast account and one API key (a secret starting with wc_live_). Dashboard → API keys → Create key → copy it. It's shown only once — save it like a password. Below, replace every wc_live_REPLACE_ME with your real key.

Two pieces: - The connection — lets the AI actually create videos through WideCast (Claude / Gemini use an "MCP server"; ChatGPT uses an "Action"). - The Skills (optional, recommended) — teach the AI to write top-tier scripts, blogs, and social posts. With Skills you get a good video, not just a video.

Step 1 — Connect WideCast

Before you connect: build the server once

WideCast isn't published to npm yet, so you run the server from this repo. One time, in a terminal:

cd /path/to/widecast/mcp-server      # the mcp-server folder in this repo
npm install

dist/index.js is ready to run. Get its absolute path — run pwd in that folder and add /dist/index.js. You'll paste that path into the config below (shown as /ABSOLUTE/PATH/TO/widecast/mcp-server/dist/index.js).

Once WideCast is on npm, you'll be able to skip this and use the one-liner "command": "npx", "args": ["-y", "@widecast/mcp-server"] instead.

Now pick your assistant.

Claude Desktop

Claude reads one settings file. You need to add a "mcpServers" section to it. **The

1 mistake is not knowing how to merge it into a file that already has other settings

— so here's exactly how.**

1. Open the file (in Claude Desktop: Settings → Developer → Edit Config, or open it directly):

macOS:    ~/Library/Application Support/Claude/claude_desktop_config.json
Windows:  %APPDATA%\Claude\claude_desktop_config.json

2. Look at what's inside. There are two cases:

Case A — the file is empty or just {}. Replace everything with this:

{
  "mcpServers": {
    "widecast": {
      "command": "node",
      "args": ["/ABSOLUTE/PATH/TO/widecast/mcp-server/dist/index.js"],
      "env": { "WIDECAST_API_KEY": "wc_live_REPLACE_ME" }
    }
  }
}

Case B — the file already has other settings (e.g. a "preferences" block, like yours). You keep everything and add "mcpServers" next to it. So if your file looks like this today:

{
  "preferences": {
    "coworkScheduledTasksEnabled": true,
    "allowAllBrowserActions": true
  }
}

…change it to this (your preferences stays exactly the same — you only add the part after the comma):

{
  "preferences": {
    "coworkScheduledTasksEnabled": true,
    "allowAllBrowserActions": true
  },
  "mcpServers": {
    "widecast": {
      "command": "node",
      "args": ["/ABSOLUTE/PATH/TO/widecast/mcp-server/dist/index.js"],
      "env": { "WIDECAST_API_KEY": "wc_live_REPLACE_ME" }
    }
  }
}

The two things people miss: 1. Add a comma right after the closing } of your existing block (the }, after preferences above). 2. Don't delete your existing settings — leave everything inside preferences untouched. You're only adding the mcpServers block.

If unsure, paste your finished file into jsonlint.com — it flags a missing comma or brace instantly.

3. Save the file, then fully quit and reopen Claude Desktop. You should see a tools / 🔌 icon — WideCast is connected.

(Requires Node.js. If npx isn't found, install it free from nodejs.org and restart.)

Cursor

Same idea, file ~/.cursor/mcp.json — add the "widecast" block inside "mcpServers" (merge the same way as above if the file already has entries):

{
  "mcpServers": {
    "widecast": {
      "command": "node",
      "args": ["/ABSOLUTE/PATH/TO/widecast/mcp-server/dist/index.js"],
      "env": { "WIDECAST_API_KEY": "wc_live_REPLACE_ME" }
    }
  }
}

Claude Code

One command — no file editing:

claude mcp add widecast node /ABSOLUTE/PATH/TO/widecast/mcp-server/dist/index.js --env WIDECAST_API_KEY=wc_live_REPLACE_ME

Gemini (Gemini CLI)

The Gemini CLI supports MCP servers via ~/.gemini/settings.json. Add (or merge, same comma rule as Claude above) the "mcpServers" block:

{
  "mcpServers": {
    "widecast": {
      "command": "node",
      "args": ["/ABSOLUTE/PATH/TO/widecast/mcp-server/dist/index.js"],
      "env": { "WIDECAST_API_KEY": "wc_live_REPLACE_ME" }
    }
  }
}

Restart the Gemini CLI; it can now call WideCast. (The Gemini web app doesn't run local MCP servers — use the Gemini CLI for tools. For the web app, paste a Skill as instructions and ask it to call the WideCast API.)

ChatGPT (Custom GPT)

ChatGPT connects via a GPT Action (WideCast already has an OpenAPI spec). Requires ChatGPT Plus.

1. ChatGPT → Explore GPTs → Create. Name it "WideCast Video Maker".

2. Configure → Actions → Create new action. Under Schema → Import from URL, paste:

https://widecast.ai/app/dashboard2/openapi.yaml

3. Under Authentication, choose API Key → Bearer, and paste your key:

wc_live_REPLACE_ME

Save the action.

Step 2 — Add the writing Skills (recommended)

The three Skills live under widecast/skills/ — video-script-writing/, blog-writing/, social-post-writing/ (each folder has a SKILL.md).

Claude (claude.ai, paid plans): Settings → Skills → Upload skill → upload each folder. (Claude Code: copy them into .claude/skills/.)
Gemini CLI: paste the contents of each SKILL.md into a GEMINI.md file in your project (Gemini CLI's instruction file).
ChatGPT: paste each SKILL.md into the GPT's Instructions box, prefaced with:

Follow this guide whenever the user wants a video script. After the user approves the
script, call the create_video action and poll get_status until it is done.

No Skills feature in your plan? Paste the SKILL.md text into a Project's custom instructions (Claude) or the chat — you'll still get most of the benefit.

Try it

Ask your assistant:

Make me a short video about why parents should let their teen get a driver's license
at 16. Render the MP4.

It writes a hook-first script → calls create_video → polls while WideCast renders → gives you the review link (and MP4 link for video output). Also try:

"Turn this YouTube link into a video" → source=video_url
"Repurpose this blog post into a 90-second video" → source=blog
"Just write the script first, let me review it" → output_type=text

Troubleshooting

Symptom	Fix
Tools / Action don't appear	Fully restart Claude Desktop / the Gemini CLI; re-import the schema in ChatGPT.
Claude won't load the config	A JSON typo — usually a missing comma or brace. Paste the file into jsonlint.com.
`npx` not found	Install Node.js from nodejs.org, then restart.
401 / `missing_api_key` / `invalid_api_key`	Key missing, mistyped, or revoked. Recreate it and update the config.
`media_too_long` / `file_too_large`	Media must be ≤ 2 minutes; uploads ≤ 100 MB.
Video stuck "processing"	Rendering takes a few minutes — ask "is it ready?" and it re-checks.

Your key bills your WideCast credits and lives only in your own config / GPT. Never share a wc_live_ key publicly or paste it into a chat message.