Particle's Radar turns 130,000 podcasts into searchable data for AI agents

Particle, the startup behind an AI-powered news reading app, has launched Radar, a search engine that turns podcast audio into structured, queryable data. According to TechCrunch, the company unveiled the product on Wednesday and positions it not merely as a transcription service but as a layer of understanding: Radar parses what is said in episodes, pulls out key quotes and highlights, and identifies the people, companies and topics under discussion.

The launch is a strategic pivot. Radar grew out of a well-liked feature in Particle's news app, which used the company's own API to surface podcast clips alongside related news stories. As the AI agent ecosystem took shape, the team concluded the capability was underused inside a consumer reader, TechCrunch reports, and decided to build it out as a standalone product.

Scale of the index

Particle says Radar transcribes more than 130,000 podcasts, which the company claims makes it the largest transcribed podcast service of its kind. Coverage includes every show in Apple's Top 200 across 135 categories, and roughly 20,000 new episodes are added to the index each day.

Transcripts come with speaker labels and rich metadata. The system performs entity recognition across people, companies, brands, products and topics, and can track how those entities are mentioned across shows over time.

Alerts, clips and ad intelligence

Users can configure alerts that fire when a tracked entity comes up, delivered either at the moment of the mention or as daily and weekly digests via email, Slack or webhook. Filters narrow the triggers, so an alert can be limited to occasions when a specific guest discusses a specific topic, or to top-ranked podcasts only.

Radar also extracts timestamped, self-contained clips, letting users listen to or read a particular remark without sitting through a full episode. Co-founder and CEO Sara Beykpour told TechCrunch that pre-selected notable clips offer a middle ground for listeners who lack time for a whole episode but do not want a summary.

Beyond search, the platform tracks ratings and reviews, chart rankings, audience size estimates, sponsorship data, brand suitability and political bias analysis. A dedicated ads search engine can surface every episode in which a given company advertises and chart how that presence trends over time, a capability with obvious appeal for competitive intelligence.

Who is paying

The early commercial traction is telling. Beykpour told TechCrunch that hedge funds are the heaviest users integrating directly with the API, drawn to data their own automated systems cannot otherwise access. AI search platforms and data resellers rank among the other top-paying customers, and Exa, a search API provider for AI agents, is one of Radar's partners. Journalists and researchers are seen as another natural audience.

All of this is available through a web interface, but Particle frames the API and its MCP support, the protocol that lets AI agents call external tools, as the real product. Pricing starts at $29 per seat per month, with a $399-per-month business tier covering 20 seats and custom pricing for API users. Particle plans to extend the service beyond podcasts to other audio, including YouTube videos and news clips.

Why it matters

Most agent and search infrastructure crawls text; audio remains largely invisible to automated systems unless someone transcribes and structures it first. Radar is a concrete, commercial example of that missing layer being built out, and its MCP support means agents can query podcast intelligence directly rather than through a human intermediary. The hedge fund interest also shows how quickly machine-readable media data becomes a trading and intelligence asset, and how spoken content, once parsed, becomes as searchable and as monetizable as the written web.