Developers/MCP Server

ContentFlow MCP Server

Connect AI agents to 50,000+ hours of podcast intelligence via the Model Context Protocol. Search content, retrieve transcripts, explore the knowledge graph, and submit new content - all from your AI assistant.

What is MCP?

The Model Context Protocol (MCP) is an open standard for connecting AI assistants to external data sources and tools. Created by Anthropic and adopted by OpenAI, Google, and others, it provides a universal interface for AI agents to search, query, and interact with services. ContentFlow's MCP server exposes podcast intelligence through 24+ tools that AI agents can call during conversations.

Setup guides

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "contentflow": {
      "type": "streamable-http",
      "url": "https://api.gocontentflow.com/mcp/"
    }
  }
}

On first use, Claude will open your browser for OAuth sign-in. After authenticating, the connection persists across sessions.

Cursor

Add to your .cursor/mcp.json in your project root:

Custom MCP client

The server uses OAuth 2.1 with browser-based authorization. Connect to the Streamable HTTP endpoint:

• Server URL: https://api.gocontentflow.com/mcp/
• Transport: Streamable HTTP (stateless)
• Auth: OAuth 2.1 Authorization Server (browser-based login via Google or email OTP)
• Scopes: read, write

Available tools

Tools are organized into categories. Each tool returns JSON and includes a hint field guiding the agent to logical next steps - enabling progressive disclosure across multi-turn research workflows.

Search

search_content

Hybrid keyword + semantic search across all content with facet filtering

search_semantic

Pure meaning-based search using vector embeddings with reranking

search_speakers

Find speakers by name or company across all podcasts

Content

get_transcript_overview

Compact transcript metadata with speaker list and segment count

get_transcript_page

Paginated transcript access (50 segments per page)

get_transcript_by_timerange

Read transcript for a specific time window

get_summaries

AI-generated summaries with citation references

get_video_details

Video metadata: title, channel, duration, speakers

Research

research_entity

Deep lookup: entity details + top mentions + related content

research_topic

Deep lookup: topic hierarchy + top videos + related topics

research_speaker

Deep lookup: speaker profile + recent appearances

get_video_profile

Complete video intelligence card with summary, topics, entities, speakers

Knowledge Graph

get_entity / search_entities

Look up companies, products, people by name or ID

get_topic / search_topics

Browse hierarchical topic taxonomy with industry verticals

get_speaker

Canonical speaker profiles with cross-episode appearances

get_video_entities / topics / sponsors

All entities, topics, or sponsors mentioned in a specific video

Discovery

get_recent_content

Latest processed videos across the platform

list_channels / get_channel_videos

Browse monitored podcast channels and their content

find_similar_segments

Vector similarity search for related discussions across videos

Write

submit_transcription_job

Submit a YouTube URL for transcription and analysis

subscribe_to_channel

Subscribe to a channel for automatic transcription of new videos

Example workflows

MCP tools are designed for progressive disclosure. Agents start with broad search, then drill into specific content. Each tool response includes hints that guide the agent to logical next steps, keeping token usage efficient across multi-turn research.

Research a company across podcasts

1

search_content(query="OpenAI fundraising strategy")

Returns ranked videos with matched entities and text chunks

2

get_video_profile(job_id from top result)

Video title, speakers, summary excerpt, top topics and entities

3

get_transcript_by_timerange(start/stop from matched segment)

Full conversation context around the relevant discussion

Find what experts say about a topic

1

research_topic("product-market fit")

Topic hierarchy, 5 top videos, related subtopics, mention count

2

get_summaries(job_id from an interesting video)

AI summary with citation-backed claims linked to source segments

3

get_transcript_by_timerange(timestamp from a citation)

Exact speaker quotes in context

Monitor a podcast channel

1

subscribe_to_channel("@a16z")

Subscription confirmed - new videos auto-transcribed

2

get_recent_content(limit=5)

Latest processed episodes with titles and durations

3

get_video_profile(job_id of latest episode)

Full intelligence card for the new episode

FAQ

What is the Model Context Protocol (MCP)?

MCP is an open standard created by Anthropic for connecting AI assistants to external data sources and tools. It lets AI agents like Claude search, query, and interact with services through a standardized protocol. Learn more at modelcontextprotocol.io.

How do I connect ContentFlow to Claude Desktop?

Add the ContentFlow MCP server URL to your Claude Desktop configuration file. The server uses OAuth 2.1 for authentication - you'll be prompted to sign in via your browser on first connection.

What data can AI agents access?

Agents can search across 50,000+ hours of podcast content, retrieve diarized transcripts with speaker labels, access AI summaries with citations, query the knowledge graph (entities, topics, speakers), and submit new content for transcription.

How does authentication work?

The MCP server uses OAuth 2.1 with browser-based authorization. On first connection, your MCP client opens a browser window for sign-in (Google OAuth or email OTP). Tokens are persisted so you don't need to re-authenticate each session.

Related

YouTube MCP Server

Production MCP server with 24 tools for Claude Desktop and Cursor - semantic search, knowledge graph, speaker diarization