Sound design coming soon.
HADEF is an AI-powered research system that helps podcast hosts book better guests by producing complete, ready-to-send episode pitches. It runs a five-agent pipeline that ingests a podcast's full episode catalog, analyzes transcripts and guest history, selects a guest based on fit (not fame), develops an original episode concept, and assembles a polished pitch package. The output is a complete episode pack: a title that matches the host's naming pattern, a premise the guest would actually want to participate in, five interview questions with preparation hints, and a hook quote pulled from the guest's own prior appearances.
What separates HADEF from topic-matching tools is what we call the golden thread. Every pitch must align three data points: a passing mention from the guest's prior interviews or writing (something they said that never got fully explored), a gap in the host's episode catalog (an adjacent topic they've never covered from this specific angle), and a timing trigger that makes the outreach feel natural rather than forced. This triangulation means the host receives an episode concept they couldn't have designed without the research, and the guest receives an invitation to expand on something they clearly care about but haven't fully articulated.
HADEF is built specifically for interview-style B2B podcasts. It is not a podcast directory, a guest marketplace, or a scheduling tool. It is the decision and synthesis layer that turns raw research into pitches worth sending.
HADEF runs a multi-agent pipeline where each stage builds on the last. First, the system ingests a podcast by pulling its full RSS feed, episode catalog, and available transcripts. It maps the show's topic coverage, identifies which guests have appeared and what they discussed, and builds a profile of the host's tone, preferences, and content gaps. This is not a surface-level keyword scan. The system reads transcripts to understand what the show actually covers versus what the episode titles suggest.
Next, HADEF selects a guest based on fit to the show's specific needs. It looks for someone with the right expertise, a track record of strong podcast appearances, and a point of view that complements (rather than duplicates) what the host has already explored. The system then searches the guest's prior interviews, articles, and public commentary for a passing mention: a quote, position, or idea that was touched on briefly but never unpacked in depth. That passing mention becomes the seed of the episode concept.
The concept development stage threads the needle between three things: a topic the guest genuinely wants to discuss (because it's rooted in their own words), a gap in the host's catalog (something adjacent to existing episodes but from a fresh angle), and a timing signal that makes the outreach feel relevant right now. The final output is a complete pitch package with a tailored episode title, a premise paragraph, five questions with interviewer preparation hints, and the original source quote that anchors the whole concept. The host receives proof of research, not a form letter.
HADEF is designed for hosts and in-house producers of interview-style B2B podcasts. These are shows where a host sits down with a guest and has a substantive conversation about business, leadership, technology, communications, or professional practice. The ideal user is running a show with under 100 episodes, actively looking for higher-caliber guests, and spending real hours each week on manual research and outreach. HADEF serves three distinct show formats: company-driven podcasts (a corporate brand in the title, often with rotating hosts and industry-aligned topics), founder-driven podcasts (the founder's name and personal brand are the show, solo host, time is scarce), and theme-driven podcasts (an abstract or conceptual title with a specific thesis the show explores through each guest).
HADEF is explicitly not built for PR agencies, podcast booking agencies, or media placement firms. Those businesses monetize the research hours that HADEF compresses, and their incentive structure is fundamentally different from a host who just wants better episodes. HADEF is also not built for solo commentary shows, panel-only formats, or news roundup podcasts, because those formats don't center on guest interviews and the system's research architecture wouldn't apply.
The addressable market is roughly 35,000 to 40,000 interview-style B2B podcasts globally. Within that, HADEF targets hosts who have already experienced the frustration of generic pitches filling their inbox, mediocre guests wasting recording slots, and high-value targets ignoring cold outreach. If you have ever received a pitch that clearly wasn't written by someone who listened to your show, HADEF exists to be the opposite of that experience.
When HADEF describes itself as context-aware, it means the system researches your specific show before it does anything else. It reads your episode catalog. It analyzes your transcripts. It maps which topics you've covered in depth, which you've only touched on, and which adjacent areas remain unexplored. It studies your guest history to understand the seniority level, professional background, and conversation style of the people who have appeared on your show. It identifies your naming patterns for episode titles and the structural conventions you follow when framing a premise.
This depth of context is what makes HADEF's output feel like it was written by someone who actually listens to your podcast, because the system did. A context-aware pitch doesn't just match a guest's expertise to your show's topic. It identifies the specific angle that belongs on your show and no one else's. Two podcasts covering the same broad subject will receive completely different episode concepts from the same guest, because the gap in each host's catalog is different, and the angle that complements their existing body of work is different.
Context-awareness also extends to the guest side. HADEF doesn't just pull a bio and a headshot. It reads what the guest has said in other interviews and articles, looking for ideas they planted but didn't water. The system looks for the passing mention that reveals what the guest actually thinks about when nobody's pressing them on it. That depth on both sides of the conversation is what allows the system to produce a pitch that feels like a natural fit rather than a cold match.
Most tools in the podcast space sit at either end of the workflow without addressing the middle. Podcast directories like Podchaser, Listen Notes, and Apple Podcasts help you discover shows and guests. Scheduling and coordination tools help you book and record. But the critical work between discovery and delivery, the research, concept development, and pitch writing that determines whether a high-value guest says yes, is still done manually or outsourced to agencies billing by the hour.
HADEF occupies that middle layer. It is the decision and synthesis system that takes raw data (episodes, transcripts, guest appearances, public commentary) and produces a pitch worth sending. Agencies charge billable hours for this research, and their output quality depends entirely on the individual researcher assigned to the task. Generic AI tools like ChatGPT can draft a pitch, but they cannot ingest your full episode catalog, cross-reference it against a guest's prior interview appearances, or verify that the proposed angle hasn't already been covered from this direction. HADEF's pipeline is podcast-specific from the ground up, not a general-purpose AI prompted to think about podcasts.
The practical difference shows up in what the host receives. A typical booking pitch says "this person is relevant to your topic and available for interviews." A HADEF pitch says "this person said something in a conversation last quarter that your audience needs to hear, you've covered the adjacent territory but never from this angle, and here are five questions that would draw out the insight your listeners are missing." That level of specificity is what turns a cold pitch into an episode concept a host would steal even if they never subscribed.
MCP stands for Model Context Protocol, an open standard that allows AI assistants to use external tools and services directly within a conversation. In practical terms, it means an AI agent like Claude, ChatGPT, or a custom assistant can call HADEF's research pipeline as a tool, the same way it might search the web or check a calendar. The host doesn't need to leave their AI workflow to use HADEF. They describe what they're looking for, the AI calls HADEF, and the results come back inside the same conversation.
HADEF exposes six tools through its MCP server: research a prospective guest, check pipeline progress, retrieve research results, retrieve a completed pitch, list all prospects in the system, and approve a pitch for sending. Each tool maps to a stage in the pipeline, so an AI agent can orchestrate the full workflow from research to delivery without the host ever opening a separate dashboard. The host stays in their preferred AI interface while HADEF does the heavy lifting in the background.
This matters because the way people interact with software is shifting. Instead of logging into a dashboard, navigating menus, and clicking buttons, a growing number of professionals are delegating tasks to AI assistants that handle the tool-switching for them. MCP integration means HADEF is ready for that shift. A host can say "find me a guest for next month's episode on supply chain resilience" to their AI assistant, and the assistant can trigger HADEF's full research pipeline, retrieve the completed pitch, and present it, all within one conversation thread.
An AI agent connects to HADEF through the MCP server using an API key for authentication. Once connected, the agent has access to the same six tools that power the HADEF dashboard: starting research on a prospect, checking the status of a running pipeline job, pulling research results, retrieving a finished pitch, listing all prospects, and approving a pitch for delivery. The agent can chain these tools together in a single workflow. For example, it can start research, poll for completion, retrieve the pitch, and present it to the user without any manual intervention between steps.
The token-based access model means each API call consumes a measured amount of compute relative to the complexity of the request. A status check is lightweight. A full research run, which involves transcript analysis, guest selection, concept development with golden thread triangulation, and pitch assembly, is a heavier operation. This usage-based structure means hosts pay for the research they actually consume rather than a flat rate that doesn't reflect how much work the system is doing on their behalf.
For developers building custom AI workflows, HADEF's MCP integration means podcast guest research becomes a callable function rather than a manual process. A production company running five shows could build an agent that researches guests for all five, compares the results, and presents the strongest options across the portfolio. A solo host could add HADEF to their weekly planning assistant so fresh episode concepts appear alongside their calendar and task list every Monday morning. The pipeline runs the same way regardless of whether a human clicked a button or an agent made the call.