Radar opens audio content to AI agents by indexing 130,000 podcasts and linking them to Model Context Protocol

Audio recordings and podcast episodes have long been a blind spot for AI agents. While these autonomous software systems can scan and analyse billions of written web texts within seconds, thousands of hours of interview discussions and analyses have remained trapped inside audio files that are difficult to retrieve programmatically without prior transcription. This gap prompted Particle, a startup founded by former Twitter engineers, to pivot entirely by launching an advanced audio search engine called Radar, designed specifically to make spoken conversations legible and programmatically accessible to AI agents and investment funds.

The engine indexes and transcribes more than 130,000 podcasts, including Apple's top 200 shows across 135 specialist categories, while adding roughly 20,000 new episodes to the index daily. Processing extends beyond speech-to-text conversion to entity extraction, speaker diarisation, timestamped highlight clipping, and the tracking of companies, brands, and products, alongside sentiment signals, bias detection, and brand suitability analytics for sponsorships.

The real shift lies in moving audio content from human listening interfaces to the structured data layer that feeds autonomous agent systems.For this reason, the company released the service not only through a conventional web interface, but via APIs and the Model Context Protocol (MCP), the standard enabling agents to connect directly to external knowledge sources. Hedge funds lead the tier of major clients integrating these programmatic interfaces to detect early investment signals, alongside agent-focused search platforms such as Exa and data distributors.

The platform also includes a dedicated engine for tracking in-episode advertisements, revealing where and how companies promote products over time alongside audience size estimates and rating data. Subscribers can configure custom alerts via email, Slack, or webhooks to receive notifications whenever a specific guest appears or a given topic is discussed, with future plans to expand indexing to YouTube videos and news broadcasts.

This transition is reshaping commercial and competitive intelligence workflows across the Gulf, Egypt, and the Levant. For regional investment teams, family offices, and asset managers, it makes incorporating unstructured data sources into internal agent pipelines necessary, as relying solely on press reports and official filings misses cues and commentary shared by chief executives and experts during long-form podcast discussions. Regional marketing teams gain the ability to audit competitors' audio ad spend globally and locally, while engineering requirements shift toward building dedicated MCP connectors between enterprise models and audio inference engines rather than standing up costly transcription infrastructure from scratch.

The push to make audio programmatically accessible shifts the focus away from human-facing search engines toward digital infrastructure built primarily for software agents, where value is measured by how quickly a quote can be retrieved and synthesised into an investment decision or executive briefing.