What is Audioscrape?
Audioscrape is a platform that turns audio into a searchable memory for AI agents. It ingests audio from multiple sources—your own calls, meetings, recordings, and a continuously growing public corpus of over 18 million podcast episodes—transcribes it with speaker diarization, and indexes every segment with timestamps and speaker attribution. AI agents can then query this index via a REST API or native MCP integration to retrieve ranked, cited segments.
The platform provides private workspaces with row-level security, ensuring your uploaded audio remains confidential and is never used to train models. It supports entity- and topic-aware search, multi-query sweeps, and offers both free and paid tiers. Audioscrape is designed for developers building agentic workflows, researchers, and enterprises needing to ground AI responses in verifiable spoken content.
Features
- Audio Ingestion & Transcription: Upload audio files or point to URLs; state-of-the-art transcription with speaker diarization and entity linking.
- Public Corpus Search: Access to over 18 million episodes from podcasts, interviews, and talks, continuously ingested and searchable.
- Semantic & Keyword Retrieval: Ranked search across both private and public audio, returning segments with speaker, timestamp, and source.
- MCP & REST API Integration: Native MCP server and REST API for easy integration with AI agents like Claude and Cursor.
- Speaker Attribution: Pin quotes to named speakers with timestamps and source links for verification.
- Multi-Query Sweeps: Agents can fan out related queries per entity or keyword and synthesize results.
Use Cases
- Ground AI agent answers in verifiable audio sources
- Search internal sales calls for key phrases and insights
- Monitor brand mentions across public podcasts
- Research topics across a vast library of spoken content
- Build agentic workflows that retrieve cited audio quotes
FAQs
-
Is my uploaded audio private?
Yes. Uploads live in your workspace, isolated by row-level security enforced in the database itself. Only you, your team, and agents holding your keys can search them. -
Do you train models on my data?
No. Your audio and transcripts are never used to train models. They exist to answer your queries, nothing else. -
How accurate are the transcriptions?
We use WhisperX with speaker diarization. Accuracy is approximately 95–99% for clear English audio, lower for heavy accents, overlapping speech, or noisy recordings. -
How do agents connect?
A native MCP server plus a REST API; one key works for both. Claude, Cursor, ChatGPT, or anything that speaks MCP or HTTP.
