Audioscrape

Searches across 1M+ hours of indexed podcast, interview, and talk audio content using dual search modes: text-based keyword matching for exact phrase discovery and semantic search for conceptual relevance. Returns segment-level results with speaker identification, precise timestamps (HH:MM:SS format), and relevance scoring (0-1 float). Implements pagination via offset/limit parameters (max 200 results per query) and supports sorting by relevance, publication date, or episode title. Results include direct URLs with timestamp anchors enabling one-click navigation to specific moments in audio.

Lists recently published podcast episodes with configurable lookback window (1-365 days, default 7 days) and optional filtering by specific podcast IDs. Returns structured episode metadata including title, podcast name, publication date (YYYY-MM-DD), duration in seconds, and direct episode URLs. Supports pagination via limit parameter (1-100 episodes per request). Designed as a lightweight alternative to full search for discovering fresh content within a time window.

Retrieves complete episode content including full transcript, metadata (title, podcast, publication date, duration), and speaker information for a specified episode ID. Enables downstream processing of full episode context rather than segment-level search results. Implementation details are partially documented; full transcript retrieval mechanism and context window handling are not fully specified in available documentation.

Exposes Audioscrape's audio search and retrieval capabilities as standardized MCP (Model Context Protocol) tools, enabling Claude, other LLM-based assistants, and AI agents to call audio search functions natively without custom API integration code. Implements OAuth 2.0 authentication with dynamic client registration following MCP spec 6/18. All tools are read-only (no mutation capabilities). Server endpoint is mcp.audioscrape.com, supporting remote MCP connections from any MCP-compatible client.

Implements tiered subscription plans (Free, Basic, Pro, Enterprise) with explicit monthly quotas for searches, API calls, and transcription minutes. Free plan: 10 searches/month, 50 transcription minutes/month. Basic plan: 50 searches/month, 50 API calls/month, 1000 transcription minutes/month. Pro plan: unlimited searches, 1000 API calls/month, 5000 transcription minutes/month. Enterprise: unlimited access. Rate limiting is enforced server-side at the MCP endpoint; quota consumption is tracked per API key and reset monthly.

Enables users to upload private audio files (meetings, calls, proprietary recordings) for indexing and search within their own Audioscrape account. Uploaded audio is transcribed, speaker-identified, and indexed using the same pipeline as public podcasts, making it searchable via the standard search_audio_content tool. Private uploads are isolated to the uploading user's account and not visible to other users. Transcription of private audio consumes the user's monthly transcription minute quota.

Supports filtering search results by podcast IDs, publication date range (date_from/date_to in YYYY-MM-DD format), and recency (last_week, last_month, last_year enum). Sorting options include relevance (default), publication date, and episode title, with ascending or descending order. Filters are applied server-side during search execution, reducing result set before returning to client. Pagination via offset/limit enables iterating through filtered results.

Optional include_context parameter in search_audio_content enables retrieval of surrounding audio segments adjacent to matched results, providing narrative context around search hits. When enabled, results include not just the matched segment but also preceding and following segments from the same episode, enabling AI models to understand broader context without requiring full episode retrieval. Context window size is not documented.