Radar Indexes 130,000 Podcasts for AI Agents via API and MCP

Particle, the AI newsreader startup founded by former Twitter engineers, launched Radar on August 26 — a podcast intelligence platform that transcribes and semantically indexes more than 130,000 podcasts, making spoken audio accessible to AI agents via both a REST API and a Model Context Protocol (MCP) endpoint. For developers building agent pipelines, the practical consequence is a structured, queryable layer over audio content that has until now been invisible to systems that crawl only text. As Particle CEO Sara Beykpour put it: "Agents are generally blind to audio; they can't see it unless something or someone has transcribed it."

The launch addresses a gap that pipeline architects have increasingly had to route around: unstructured audio doesn't participate in retrieval-augmented generation or agent tool calls unless it has been converted into indexed, structured data.

Corpus and Ingestion

Radar's index covers all Apple Top 200 podcasts across 135 verticals, with 20,000 new episodes added daily. Transcriptions carry speaker labels and structured metadata that identifies entities — people, companies, brands, products, and topics — at the segment level, not just the document level. This entity-resolution layer is what separates the platform from a batch transcription service: Radar can track a named entity across the entire corpus and surface every timestamped mention, enabling cross-podcast monitoring rather than single-episode search.

The platform pre-selects "notable clips" — self-contained excerpts with timestamps returnable as both text and playable audio references. A dedicated podcast-ads search engine tracks which episodes carry a given company's advertising and surfaces trend data over time. Additional signals include political bias scores, chart rankings, audience size estimates, sponsorship data, and brand suitability flags — all accessible via API.

Developer Integration

The primary integration path is the Radar API, with custom pricing negotiated per use case. The MCP endpoint extends the same corpus to agent tool-call architectures without requiring a custom REST integration for every agent framework. Exa, the AI-agent-oriented search API provider, is named as an existing Radar partner, indicating the platform is already composable with downstream retrieval layers used in production stacks.

Alert delivery supports email, Slack, and webhook targets, filterable by guest identity, topic co-occurrence, and podcast tier. Webhook delivery makes Radar a viable event source for agent pipelines that need to react to new mentions rather than poll on a schedule.

Hedge funds are currently the highest-volume direct API customers, per Beykpour — a detail that reflects the data's value for time-sensitive, entity-specific monitoring. AI search platforms and data resellers are also cited as top-paying customer categories.

Pricing

Programmatic access through the API and MCP carries separately negotiated pricing. Particle has not disclosed rate limits, latency targets, or token-level pricing for the API tier.

Roadmap

Radar's current corpus is podcast-only, but Particle has stated plans to extend indexing to YouTube videos and news audio clips. The architecture — entity extraction, timestamped segmentation, cross-corpus tracking — maps onto any audio source with sufficient metadata, making ingestion throughput the primary engineering challenge in expansion rather than pipeline redesign.

Audio is becoming a first-class data type in the agent tool ecosystem rather than a format that agents must outsource to human summarisers. Particle's pivot from a consumer news reader to a B2B API business tracks a pattern visible across the 2026 AI stack: proprietary data surfaces, not model capability alone, are increasingly the defensible layer. Platforms that expose previously opaque corpora as structured, queryable endpoints stand to become infrastructure dependencies for the agent pipelines being assembled today.