How VCs use podcast conversations for deal sourcing

In this post

• Describe the companies you invest in by sector, stage, region and timeframe, and get a shortlist where each company links to the podcast moment it came from.

• Ask follow-up questions to find patterns across the shortlist.

• Pull a founder interview and turn it into questions for a first meeting.

• Examples from stablecoins, biotech, robotics, European startups and cybersecurity.

A lot of VC teams now pull structured data straight into their AI tools. Crunchbase launched an MCP server in July, and PitchBook has an official connector for Claude. Those sources are built around records: who raised, how much and from whom.

Podcasts add a different kind of information: what people said. A founder explains why they started the company. A daily news show covers a new round. An investor talks through a category they’re watching. That’s useful for sourcing, and it’s hard to search because it’s audio.

Particle makes podcasts searchable. We actively transcribe 140,000+ podcasts, and new episodes are typically searchable within minutes of airing. You can reach them through an MCP server, so they sit next to your other sources in Claude, ChatGPT or Cursor, or through a REST API. If you’d rather search by hand, Radar is our podcast search engine in the browser, and it’s where the episode links in this post open.

The workflow is the same whatever your fund invests in. Here it is for one sector, cybersecurity:

1. Describe the companies you want

Connect Particle to your MCP client (see “Set it up” below), then describe the companies you’re looking for in plain language. Good prompts cover five things:

• Sector or problem: “stablecoin infrastructure,” “AI for drug discovery.”

• Stage: “seed or Series A.”

• Timeframe: “recent,” or “since July.”

• Type of conversation: founder interviews tell you how the people building a company think. News segments tell you who raised.

• A name, when you have one: Particle resolves it to a company or a person in its knowledge graph before it searches, so "Chao Cao" returns that founder's interview rather than everything that sounds like the name, and a line a transcript spells "Stryker" still resolves to Straiker.

You can add a region too. Here’s one company each of these prompts found:

Each search returned several companies at different stages, so keep the ones that fit your fund!

Example: cybersecurity

Behind the cybersecurity prompt in the video, the assistant calls three Particle tools. It searches transcripts for conversations about cybersecurity startups raising early rounds (particle_podcast_search_transcripts), resolves the companies named in them against Particle’s knowledge graph (particle_company_resolve), and pulls the source passages from each episode (particle_podcast_get_episode). Each company comes back with the show, the date and the exact line.

The results included later-stage companies like Onyx Security (Series B) and Horizon3 (Series E). Here are the six seed and Series A companies:

Each “Where it came up” link opens the episode in Radar at the line where the company is discussed. Stage and funding come from each company’s funding announcement or launch coverage.

2. Ask follow-up questions

Keep going in the same conversation. For the cybersecurity shortlist:

What themes do these companies point to?

• Agent identity and access: Oak manages identity for people, machines and AI agents in one system. Neo lets security teams set policies for an agent’s tool calls, API access and data movement.

• Runtime protection: AIR Security, Straiker and ZeroDrift watch AI systems while they work and step in when something goes wrong. ZeroDrift does it for compliance, checking each outgoing message before it’s sent.

• Defensive AI models: Corma is building a foundation model for cyber defense.

You can ask the same question about any shortlist.

3. Go deeper with a founder interview

A news segment tells you what a company does and who funded it. A founder interview tells you how the people building it think, which is what you want before a first meeting.

Here’s an example from robotics. Sancho builds general-purpose robots that connect separate machines on a manufacturing line, handling the handoffs between them that people do by hand today. It raised a $7 million seed round co-led by Fusion Fund and Catapult Ventures in July.

Ask:

Find podcast interviews with Sancho’s CEO, Chao Cao. What is the company betting on?

Particle finds Cao on The OPTIM Update, a podcast hosted by Bogdan Cristei of Optim VC. His main bet goes against the most common approach in robot AI, where models learn from camera pixels and very large datasets. Sancho builds 3D geometric models of the world instead. In his words, even “if you got a pixel level accuracy, you are not necessarily guaranteed to have physical level accuracy.” He also expects a lot of the robot data collected today to be “wasted in the future.”

That’s a specific, testable view of the market. Now ask for the next step:

Based on this interview, what should I ask Sancho’s founders in a first meeting?

Questions like these come straight out of what he said:

• Where have 3D models beaten camera-based approaches in a customer’s facility, and how do you measure that?

• Which machine handoffs are customers paying to automate first?

• What robot data does Sancho collect that you expect to stay useful as models change?

• Sancho runs its software on the robot’s own computer. What does that do to deployment cost?

Other ways VC teams use it

• Pull every line about one company or person. “What has been said about [company] on podcasts?” Particle resolves the name to an entity first and returns the mentions grouped by episode, so you get the company you meant rather than everything that sounds like it.

• Watch portfolio companies. “Alert me when [portfolio company] is mentioned on a podcast.” Each alert watches one company, person, brand or topic across the indexed catalog, and sends matches as they happen or as a daily or weekly digest, with the clip and transcript attached. Paid plans include alerts: 1 on Individual, 5 on Business.

• Track a founder’s podcast appearances. “Which podcasts has [founder name] been a guest on?” Particle keeps an appearance history for each guest, including how many shows they’ve been on and their first and most recent appearance.

• Spot a press tour. “Who is making the rounds on podcasts right now?” Particle’s trending-guest list counts only people who appeared on two or more different shows in the last 30 days, which separates a press tour from a show regular.

• See how often a topic comes up. “How many podcast episodes discussed [topic] each month this year?” Particle counts matching episodes by day, week or month.

Set it up

The quickest way to get started is to hand the setup to your assistant. In Claude or ChatGPT, paste:

Read particle.pro/agents.md and connect me to Particle over MCP.

The assistant reads that page, walks you through its own settings, and asks you to sign in when it gets there. In ChatGPT that includes turning on developer mode yourself, which needs a paid plan. If you would rather follow written steps, the setup guide has every client, with one-click installs for Cursor and VS Code.

To do it by hand, the server is https://mcp.particle.pro and you sign in through the browser the first time:

Two things that can trip people up. On Claude Team and Enterprise plans, an Owner or Primary Owner adds the connector once under Organization settings before anyone else can connect it. And OpenAI’s own pages disagree on which plans get developer mode: its guide lists Pro, Plus, Business, Enterprise and Edu, while its help center describes full MCP support on Business, Enterprise and Edu.

Prefer your own stack? The same search is one REST call, GET /v1/podcasts/episodes/search with semantic_search, since and limit, authenticated with an X-API-Key header. Create an account, generate a key, and the docs have the rest.

FAQ

What’s indexed? 140,000+ podcasts and counting, including the Apple Podcasts Top 200 in the US across 130+ English-language verticals, plus independent shows.

How current is the data? Most episodes are transcribed, speaker-labeled and searchable within minutes of publication.

What does it cost? Individual is $29 a month for 10,000 requests, one seat and one alert. Business is $399 a month for 100,000 requests, 20 seats, five alerts and the premium endpoints (ad intelligence, publisher landscape, brand safety and suitability). Enterprise is custom, and adds a real-time firehose, high request volumes with SLAs, custom integrations, access to all endpoints and dedicated support: api@particle.pro.

Is there a free trial? Yes. Picking a plan comes with $10 in free credits, about 1,000 requests, and you are not charged until you have used them.

Do the episode links work without an account? Anyone can open them and read the transcript around the highlighted line. Reading the rest of an episode needs an account.

Can we build this into our own tools? Yes. Particle’s API terms cover the permitted uses of data in internal tools, dashboards and analytics platforms, as well as external-facing applications. Certain uses are not allowed, including the resale of data, bulk downloading or storage beyond what is reasonably necessary for display, and training of machine learning models. Any attribution information that is included in API output must also be displayed to end users.

What’s Radar? Radar is Particle’s search engine for podcasts: entity, keyword and semantic search in the browser, with no code. The episode links in this post open there.

Start sourcing

Find new prospects and explore emerging themes. You start with $10 in free credits, about 1,000 requests. Start at particle.pro.