Connect your AI

What you can do

MCP is the bridge between your AI and oDarkly. Instead of clicking through the dashboard, you describe what you want and the assistant runs it: create a project, generate a script, render the MP4, queue YouTube uploads, or create and track automations. Everything runs in the cloud — nothing to install locally.

Works in Cursor, Claude Desktop, ChatGPT (MCP connectors), and any client that supports remote HTTPS servers.

1. Create your API key

The key identifies your account when the AI connects to oDarkly. It is shown only once — copy it and store it somewhere safe.

1. Sign in to your dashboard.
2. Go to Settings → Connect your AI.
3. Create the key and copy the odk_live_… value right away.
4. Want full video + YouTube automation? Also enable the generate and youtube:write scopes.

Open dashboard settings →

2. Connect your assistant

Paste the URL below into the app you use. No Node install or local script — the server is already live.

https://odarkly.com/api/mcp

Send your API key on every request:

Authorization: Bearer odk_live_YOUR_KEY_HERE

Client without remote support?

If the app only accepts local (stdio) MCP, use the mcp-remote package as a bridge — it connects to the oDarkly remote server for you:

{
  "mcpServers": {
    "odarkly": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://odarkly.com/api/mcp",
        "--header",
        "Authorization: Bearer odk_live_YOUR_KEY_HERE"
      ]
    }
  }
}

Test connection

With your odk_live_… key from the dashboard, this command should return protocolVersion and odarkly serverInfo:

curl -X POST https://odarkly.com/api/mcp \
  -H "Authorization: Bearer odk_live_YOUR_KEY_HERE" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"test","version":"1.0"}}}'

Security

Remote MCP uses the same authentication as API v1. Each user can only access their own account — there is no shared login.

• Keys are personal: create them in Settings → Connect your AI and revoke if leaked.
• You choose what each key can do: read-only, create projects, generate videos, or publish to YouTube.
• All requests must use HTTPS. Never expose your key on public sites or repositories.
• Per-key and per-IP rate limits help prevent abuse.
• During dashboard maintenance, MCP is unavailable for regular users (admins retain access).

Cursor configuration

Add to .cursor/mcp.json or Cursor Settings → MCP. Paste the remote URL and Authorization header with your odk_live_… key.

{
  "mcpServers": {
    "odarkly": {
      "url": "https://odarkly.com/api/mcp",
      "headers": {
        "Authorization": "Bearer odk_live_YOUR_KEY_HERE"
      }
    }
  }
}

Claude configuration

In Claude Desktop: Settings → Connectors → Add custom connector. URL: https://odarkly.com/api/mcp with Bearer auth and your API key. In Claude Code, use the JSON block below.

{
  "mcpServers": {
    "odarkly": {
      "url": "https://odarkly.com/api/mcp",
      "headers": {
        "Authorization": "Bearer odk_live_YOUR_KEY_HERE"
      }
    }
  }
}

ChatGPT configuration

ChatGPT only connects to remote MCP servers over HTTPS (no local stdio).

1. Open ChatGPT → Settings → Connectors (or Developer / Apps).
2. Add a custom MCP connector.
3. Server URL: https://odarkly.com/api/mcp — auth: Bearer odk_live_…

https://odarkly.com/api/mcp

From script to YouTube

With the right scopes on your key, you can ask for all of this in one conversation. The AI calls the tools in order — you do not need to memorize the names.

1. Create the project and generate script, images, and narration. create_project · generate_script · generate_scene_images · generate_audio
2. Render the video and wait until it is ready. start_render · get_render_job
3. Check your channels and queue the video for publishing. list_youtube_channels · queue_youtube_video
4. Set up a funnel to publish on a fixed schedule if you want the channel on autopilot. create_funnel · update_funnel
5. Create and track dashboard Automations: which are ready, which ran, and which already posted. create_automation · list_automations · list_automation_runs
6. Create influencer, carousel, UGC, cinema, or GUM Animator from chat — the same tools as the dashboard. create_influencer · create_gum_character · create_carousel · create_ugc_video · create_cinema_video · prepare_gum_animator · create_gum_animator

To publish on YouTube, connect your channel once under Channels in the dashboard — the AI uses the connection you already set up.

Example prompts

Copy, tweak, and send these in your assistant after connecting MCP:

• “

  Create a horror Short about an abandoned lighthouse, generate everything, and render the video.

• “

  List my ready projects and queue the latest one on my main channel.

• “

  Create a funnel on channel X posting Mon–Fri at 6pm with review before upload.

• “

  List my automations, tell me which are ready and whether the last run posted to YouTube. Then create a horror Short automation through publish.

• “

  Create a female influencer, 25, curly hair, casual look. Then a 7-slide Instagram carousel about a skincare routine with her.

• “

  List my GUM Creator sheets. Prepare this MP4 in GUM Animator and recast the main person as the latest sheet.

Tool reference

These are the commands your AI can run for you. Each one needs the scope shown on the right — without it, the tool will not appear in the list.

REST API v1

All endpoints live under /api/v1/. Send JSON for POST requests with Content-Type: application/json. Authenticate with a Bearer token.

Authentication

Include your API key in every request:

Authorization: Bearer odk_live_YOUR_KEY_HERE

API key scopes

Enable only what you need. Default keys can read and create projects. To produce and publish videos through your AI, turn on the two extra scopes below.

• read View projects, automations, credits, profile, and render progress
• projects:write Create projects and manage automations
• generate Generate script, images, audio, and render the MP4
• youtube:write Publish to YouTube and manage channel funnels

Rate limits

Limits apply per API key. When exceeded, the API returns HTTP 429.

• 300 read requests per minute per API key
• 30 write requests (POST) per minute per API key

Example requests

Replace odk_live_YOUR_KEY_HERE with your actual key. Never commit keys to version control.

curl -H "Authorization: Bearer odk_live_YOUR_KEY_HERE" \
  https://odarkly.com/api/v1/me.php

curl -H "Authorization: Bearer odk_live_YOUR_KEY_HERE" \
  https://odarkly.com/api/v1/credits.php

curl -H "Authorization: Bearer odk_live_YOUR_KEY_HERE" \
  "https://odarkly.com/api/v1/projects.php?limit=10"

curl -X POST \
  -H "Authorization: Bearer odk_live_YOUR_KEY_HERE" \
  -H "Content-Type: application/json" \
  -d '{"type":"SHORT","topic":"A haunted lighthouse","niche":"TERROR"}' \
  https://odarkly.com/api/v1/projects.php