MCP (Model Context Protocol)

Use TranscriptAPI to fetch transcripts, search YouTube, browse channels, and list playlists — all from within your AI assistant via Model Context Protocol (MCP).

Claude Tutorial

Complete guide to integrate TranscriptAPI with Claude AI assistant via MCP.

RecommendedChatGPT Tutorial

Step-by-step guide to connect TranscriptAPI with ChatGPT using OAuth authentication.

Agent Builder Tutorial

Learn how to set up TranscriptAPI with OpenAI Agent Builder for AI workflows.

What is MCP?

Model Context Protocol (MCP) is a standardized framework that enables AI assistants like Claude and ChatGPT to connect directly with external tools and data sources. With TranscriptAPI’s MCP integration, your AI can fetch YouTube transcripts on-demand without manual copy-pasting or switching between applications.

Key Benefits:

• Direct integration into your AI workflow
• No manual transcript fetching needed
• Consistent, structured data format
• Works with any MCP-compatible platform

Why Use TranscriptAPI with MCP?

TranscriptAPI offers unique advantages for YouTube Transcript MCP integration:

Universal Compatibility

TranscriptAPI works with any platform that supports MCP. Whether your platform uses OAuth for seamless authentication (like Claude and ChatGPT) or requires manual API key configuration (like OpenAI Agent Builder), we provide both options to ensure maximum compatibility.

Proven Reliability

• Battle-tested with major AI platforms
• Consistent performance across all authentication methods
• Same powerful transcript extraction engine as our REST API
• Flexible authentication options (OAuth or API Key) based on your platform’s requirements

Prerequisites

Before setting up TranscriptAPI with MCP, ensure you have:

• A TranscriptAPI account (sign up free)
• An AI assistant or platform that supports MCP
• For OAuth: A web browser for the initial authorization
• For API Key: An active API key from your dashboard

MCP Server Configuration

Get your MCP configuration details from the MCP dashboard or use these settings:

Property	Value
Server Name	Transcript API
Server URL	https://transcriptapi.com/mcp
Description	Use TranscriptAPI to fetch YouTube video transcript automatically
Icon	Download Icon

Quick Setup

Visit your MCP dashboard for a personalized setup guide with these configuration details and authentication options tailored to your chosen platform.

Authentication

TranscriptAPI supports two authentication methods to work with different MCP platforms:

OAuth (Recommended)

OAuth Authentication

OAuth provides automatic, secure authentication without manual key management. Supported by:

• Claude (Dynamic Client Registration - optional credentials)
• ChatGPT (Static Client Registration - requires Client ID/Secret)
• Any MCP client supporting OAuth 2.1

Two Registration Methods:

TranscriptAPI supports both Dynamic Client Registration (DCR) and Static Client Registration to work with different platforms:

1. Dynamic Client Registration (Claude, optional)

• Claude and other clients that support DCR can automatically register
• No Client ID/Secret needed - just add the MCP server URL
• The client automatically discovers OAuth endpoints and registers itself
• You’ll be redirected to authorize access (one-time)

2. Static Client Registration (ChatGPT, required)

• ChatGPT requires Client ID and Client Secret upfront
• Get your credentials from the MCP Integration dashboard
• Select “OAuth (Recommended)” and click “Generate Credentials”
• Copy your Client ID and Client Secret
• Add them to your ChatGPT MCP connector configuration

How it works:

1. Add the MCP server URL to your client
2. For ChatGPT: Add Client ID and Client Secret
3. For Claude: Credentials are optional (DCR supported)
4. The client automatically discovers our OAuth endpoints
5. You’ll be redirected to authorize access (one-time)
6. Tokens are managed automatically with refresh support

Technical Details:

• Uses OAuth 2.1 with both Dynamic Client Registration (RFC 7591) and Static Registration
• Authorization server metadata: /.well-known/oauth-authorization-server
• Protected resource metadata: /.well-known/oauth-protected-resource
• Required scope: mcp:access
• Access tokens prefixed with oat_

Benefits:

• No manual key management (for DCR clients)
• Automatic token refresh
• Secure, time-limited access
• Revocable permissions
• Platform-specific support (ChatGPT and Claude)

API Key

API Key Authentication

API Key authentication provides simple, direct access for platforms that don’t support OAuth. Required for:

• OpenAI Agent Builder
• Custom MCP implementations
• Development/testing environments

How it works:

1. Get your API key from the API Keys dashboard
2. Add it to your MCP client configuration
3. The key is sent as a Bearer token with each request

Configuration:

{
  "mcpServers": {
    "transcriptapi": {
      "url": "https://transcriptapi.com/mcp",
      "apiKey": "sk_your_api_key_here"
    }
  }
}

Technical Details:

• Bearer token authentication: Authorization: Bearer sk_...
• API keys prefixed with sk_
• Keys never expire (unless manually revoked)
• Same key works for both REST API and MCP

Security Best Practices:

• Store keys in environment variables
• Never commit keys to version control
• Use separate keys for different environments
• Rotate keys regularly

Available Tools

TranscriptAPI exposes 6 tools through MCP:

Tool	Description	Cost
get_youtube_transcript	Fetch video transcript as markdown or JSON	1 credit
search_youtube	Search YouTube for videos or channels	1 credit
get_channel_latest_videos	Get latest ~15 videos from a channel (RSS)	Free
search_channel_videos	Search within a specific channel	1 credit
list_channel_videos	Paginated list of all channel uploads	1 credit/page
list_playlist_videos	Paginated list of playlist videos	1 credit/page

All channel tools accept flexible input: @handle, full YouTube URL, or UC... channel ID.

get_youtube_transcript

Extract transcripts from any YouTube video with full control over the output format.

Parameters:

Parameter	Type	Default	Description
video_url	string	required	YouTube video URL or 11-character video ID
send_metadata	boolean	true	Include video title, author, and thumbnail
format	string	"text"	Output format: "text" (markdown) or "json" (structured)
include_timestamp	boolean	true	Include timestamps with each segment

Accepted URL Formats:

• Full URL: https://www.youtube.com/watch?v=dQw4w9WgXcQ
• Short URL: https://youtu.be/dQw4w9WgXcQ
• Video ID only: dQw4w9WgXcQ

Example Usage:

// Get transcript as markdown with metadata (default)
get_youtube_transcript({
  video_url: "https://www.youtube.com/watch?v=dQw4w9WgXcQ"
})

// Returns:
{
  "content": "# Metadata\n## Title: Rick Astley - Never Gonna Give You Up\n## Author: RickAstleyVEVO\n## Author URL: https://www.youtube.com/@RickAstley\n## Thumbnail: https://i.ytimg.com/vi/dQw4w9WgXcQ/hqdefault.jpg\n\n# Transcript\n[0.0s] Never gonna give you up\n[4.12s] Never gonna let you down..."
}

// Get JSON format without metadata
get_youtube_transcript({
  video_url: "dQw4w9WgXcQ",
  format: "json",
  send_metadata: false
})

// Returns:
{
  "content": {
    "transcript": [
      {
        "text": "Never gonna give you up",
        "start": 0.0,
        "duration": 4.12
      },
      {
        "text": "Never gonna let you down",
        "start": 4.12,
        "duration": 3.85
      }
    ]
  }
}

// Get plain text without timestamps
get_youtube_transcript({
  video_url: "dQw4w9WgXcQ",
  format: "text",
  include_timestamp: false,
  send_metadata: false
})

// Returns:
{
  "content": "# Transcript\nNever gonna give you up Never gonna let you down Never gonna run around and desert you..."
}

Output Formats:

Text Format (Default)

Returns markdown-formatted text with optional metadata header and timestamped transcript lines.

# Metadata
## Title: Video Title Here
## Author: Channel Name
## Author URL: https://www.youtube.com/@ChannelName
## Thumbnail: https://i.ytimg.com/vi/VIDEO_ID/hqdefault.jpg

# Transcript
[0.0s] First line of transcript
[3.5s] Second line of transcript
[7.2s] Third line of transcript

JSON Format

Returns structured data with transcript segments and optional metadata.

{
  "transcript": [
    {
      "text": "First line of transcript",
      "start": 0.0,
      "duration": 3.5
    },
    {
      "text": "Second line of transcript",
      "start": 3.5,
      "duration": 3.7
    }
  ],
  "metadata": {
    "title": "Video Title Here",
    "author_name": "Channel Name",
    "author_url": "https://www.youtube.com/@ChannelName",
    "thumbnail_url": "https://i.ytimg.com/vi/VIDEO_ID/hqdefault.jpg"
  }
}

search_youtube

Search YouTube for videos or channels. Returns structured results with video IDs, titles, view counts, and thumbnails.

Parameters:

Parameter	Type	Default	Description
query	string	required	The search query
search_type	string	"video"	Type of search: "video" or "channel"
limit	number	10	Maximum number of results to return

Example Usage:

search_youtube({
  query: "TED talk artificial intelligence",
  search_type: "video",
  limit: 5
})

// Returns:
{
  "content": {
    "results": [
      {
        "video_id": "abc123",
        "title": "The future of AI | TED Talk",
        "channel": "TED",
        "view_count": "2.5M views",
        "published": "2024-03-15",
        "thumbnail": "https://i.ytimg.com/vi/abc123/hqdefault.jpg"
      }
      // ... more results
    ]
  }
}

get_channel_latest_videos

Get the ~15 most recent videos from a YouTube channel via RSS. Free — no credits consumed.

Parameters:

Parameter	Type	Default	Description
channel	string	required	Channel identifier: @handle, channel URL, or UC... ID

Accepted Input Formats:

• Handle: @TED
• URL: https://www.youtube.com/@TED
• Channel ID: UCsT0YIqwnpJCM-mx7-gSA4Q

Example Usage:

get_channel_latest_videos({
  channel: "@TED"
})

// Returns:
{
  "content": {
    "channel": {
      "name": "TED",
      "url": "https://www.youtube.com/channel/UCsT0YIqwnpJCM-mx7-gSA4Q"
    },
    "results": [
      {
        "video_id": "xyz789",
        "title": "The next breakthrough in AI",
        "published": "2024-06-01",
        "thumbnail": "https://i.ytimg.com/vi/xyz789/hqdefault.jpg"
      }
      // ... up to ~15 results
    ]
  }
}

search_channel_videos

Search for videos within a specific YouTube channel. Costs 1 credit per request.

Parameters:

Parameter	Type	Default	Description
channel	string	required	Channel identifier: @handle, channel URL, or UC... ID
query	string	required	Search query within the channel
limit	number	10	Maximum number of results to return

Example Usage:

search_channel_videos({
  channel: "@TED",
  query: "education",
  limit: 5
})

// Returns:
{
  "content": {
    "results": [
      {
        "video_id": "def456",
        "title": "The future of education | TED Talk",
        "view_count": "1.2M views",
        "published": "2024-01-20",
        "thumbnail": "https://i.ytimg.com/vi/def456/hqdefault.jpg"
      }
      // ... more results
    ]
  }
}

list_channel_videos

List all videos in a channel with pagination (~100 per page). Costs 1 credit per page.

Parameters:

Parameter	Type	Default	Description
channel	string	none	Channel identifier: @handle, URL, or UC... ID (first call)
continuation	string	none	Pagination token from previous response (subsequent calls)

Provide either channel (first call) or continuation (next pages) — not both.

Example Usage:

// First call — start listing
list_channel_videos({
  channel: "@TED"
})

// Returns:
{
  "content": {
    "results": [
      {
        "video_id": "ghi789",
        "title": "How AI could empower any business",
        "view_count": "800K views",
        "published": "2024-05-10"
      }
      // ... ~100 results
    ],
    "continuation_token": "Ehl...",
    "has_more": true
  }
}

// Next page — use continuation token
list_channel_videos({
  continuation: "Ehl..."
})

list_playlist_videos

List videos in a YouTube playlist with pagination (~100 per page). Costs 1 credit per page.

Parameters:

Parameter	Type	Default	Description
playlist	string	none	Playlist URL or ID starting with PL, UU, LL, FL, or OL (first call)
continuation	string	none	Pagination token from previous response (subsequent calls)

Provide either playlist (first call) or continuation (next pages) — not both.

Example Usage:

// First call — start listing
list_playlist_videos({
  playlist: "PLOGi5-fAu8bFIs_Lbp-MNwPKpPjBJGUzr"
})

// Returns:
{
  "content": {
    "results": [
      {
        "video_id": "jkl012",
        "title": "Introduction to Machine Learning",
        "channel": "TED-Ed",
        "published": "2024-02-15"
      }
      // ... more results
    ],
    "continuation_token": "Qmx...",
    "has_more": true
  }
}

// Next page — use continuation token
list_playlist_videos({
  continuation: "Qmx..."
})

Platform Integration

TranscriptAPI MCP works with any platform that supports the Model Context Protocol. Here are guides for popular platforms:

Tested Platforms

• Connect to Claude - OAuth authentication support
• Connect to ChatGPT - OAuth authentication support
• Connect to OpenAI Agent Builder - API key authentication

Universal Compatibility

TranscriptAPI works with any tool, framework, or custom implementation that supports MCP:

• OAuth-enabled platforms: Automatic authentication and token management
• API key platforms: Manual configuration with your TranscriptAPI key
• Custom implementations: Choose the authentication method that fits your needs

Future-Proof Integration

As new MCP-compatible platforms emerge, TranscriptAPI will work with them out of the box. Our standards-compliant implementation ensures long-term compatibility.

Credit Usage & Rate Limits

MCP requests consume credits and are subject to rate limits just like REST API calls:

Credit Cost

Tool	Cost
get_youtube_transcript	1 credit
search_youtube	1 credit
get_channel_latest_videos	Free
search_channel_videos	1 credit
list_channel_videos	1 credit per page
list_playlist_videos	1 credit per page

• Only charged on successful (HTTP 200) responses
• Failed requests (4xx, 5xx) don’t consume credits

Rate Limits

• 200 requests per minute per API key/OAuth token
• Rate limits apply across all API usage (REST + MCP combined)

See API Rate Limits for detailed information and best practices.

Monitor Your Usage

MCP tools may make multiple requests during a conversation. Monitor your credit usage in the dashboard to avoid unexpected consumption.

Error Handling

All MCP tools return user-friendly error messages that AI assistants can understand and communicate:

Common Errors

Error Code	Description	Retry?	Solution
401	Invalid authentication	❌	Check your API key or re-authorize OAuth
402	Insufficient credits	❌	Purchase more credits or upgrade plan
404	Transcript not available	❌	Video may not have captions or be private
408	Timeout / Retry	✅	Temporary failure - retry after 1-5 seconds
422	Invalid YouTube URL	❌	Provide a valid YouTube URL or video ID
429	Rate limit exceeded	✅	Wait before retrying (see Retry-After header)
500	Server error	⚠️	Temporary issue, contact support if persistent
503	Service unavailable	✅	Service temporarily down - retry after 1-5 seconds

Error Response Format

Errors are returned in a format that AI assistants can easily interpret:

{
  "content": "Unable to fetch transcript: Payment required. You have no credits remaining. Please purchase more credits at https://transcriptapi.com/dashboard/billing"
}

The AI assistant will communicate these errors naturally to the user.

Best Practices

Security

• OAuth platforms: Let the platform handle token storage and refresh
• API key platforms: Store keys in environment variables, never in code
• Rotate API keys regularly for enhanced security
• Monitor usage through the dashboard to detect anomalies

Performance

• Batch requests when possible to stay within rate limits
• Cache transcripts locally if you need them multiple times
• Handle 402 errors gracefully—inform users about credit status
• Implement exponential backoff for retries on 408/429/503 errors

Integration Tips

• Test with common videos first to ensure configuration works
• Provide context to the AI about video content for better responses
• Use metadata to give AI information about video source and author
• Choose the right format: JSON for structured processing, text for natural language

Development Workflow

You can develop and test with API keys for quick setup, then switch to OAuth for production deployments when supported by your platform.

Troubleshooting

OAuth Issues

• “Authorization failed”: Clear browser cookies and try again
• “Invalid redirect”: Ensure you’re using the latest MCP client version
• Token expiration: OAuth tokens auto-refresh; if issues persist, re-authorize

API Key Issues

• “Invalid API key”: Verify key starts with sk_ and is active in dashboard
• “No credits”: Check credit balance at /dashboard/billing
• Key not working: Ensure no extra spaces when copying the key

General Issues

• No transcript returned: Video might not have captions available
• Slow responses: Check if video is very long (>2 hours)
• Connection errors: Verify your internet connection and firewall settings

Next Steps

Ready to get started? Here’s what to do next:

1. Get your credentials: Sign up and get an API key or prepare for OAuth
2. Choose your platform: Select from our platform-specific guides
3. Configure MCP: Follow the configuration examples above
4. Test it out: Try fetching a transcript from a YouTube video
5. Explore the API: Check our REST API Reference for advanced usage

Need live testing? Try our Swagger UI to experiment with the API directly.

Related

API Reference Complete REST API documentation with examples and parameters.

Getting Started Quick start guide for TranscriptAPI basics and setup.

Contact Support Get help with technical issues or account questions.