Bridging the Audio Gap for AI Agents
Most AI agents and search platforms are optimized for text, leaving them effectively blind to the vast amount of information contained in audio. Radar, a tool developed by the startup Particle, addresses this by providing a comprehensive indexing and transcription layer for podcasts. By transcribing over 130,000 podcasts—including the Apple Top 200—and adding 20,000 episodes daily, Radar transforms unstructured audio into a searchable, machine-readable format.
Intelligence and Programmability
Beyond simple transcription, Radar extracts rich metadata, including speaker labels and entity recognition (people, brands, products, and topics). This intelligence allows for several high-value use cases:
- Entity Tracking & Alerts: Users can monitor specific topics or guests across the entire podcast ecosystem, receiving alerts via email, Slack, or webhooks when mentions occur.
- Clip Extraction: The system identifies and extracts self-contained, timestamped highlights, allowing users to consume key moments without listening to full episodes.
- Ad Intelligence: A dedicated search engine tracks where companies advertise, enabling analysis of sponsorship trends and brand suitability over time.
API-First Architecture
While Radar offers a web interface, its primary value lies in its API and Model Context Protocol (MCP) support. This allows developers to integrate podcast intelligence directly into AI agents and data platforms. The service has already seen significant adoption from hedge funds seeking proprietary data insights, as well as AI search platforms and data resellers. Future development plans include expanding support beyond podcasts to include YouTube videos and other news-based audio clips.