ForgeHouse: turning expert mentors into AI agents available 24/7
Two weeks ago I noticed I kept paying for mentorship calls where 80% of the value was the mentor's thinking patterns, not the live interaction. Same frameworks, same diagnostic questions, same instincts. Every time.
So I built ForgeHouse.
The idea: extract how a real mentor actually thinks, not surface-level advice, but the decision trees, the frameworks, the questions they ask before giving an answer. Turn that into an AI agent anyone can talk to at any hour.
Day 1 (Feb 20): Domain, landing page, chat system, extraction pipeline. Pitched the founder of an established mentoring platform. He told me it would cannibalize his business. 30 minutes later he was debating product framing with me.
Week 1: First mentor extracted. Colin Chapman, 26 years in GTM and outbound sales. 45 extraction exchanges pulling out how he actually thinks. 23 calibration rounds where he stress-tested the agent against real scenarios. His reaction: the extraction itself was valuable because it forced him to articulate things he'd been doing on autopilot for decades.
Week 2: Shipped retention features. Context memory (agent remembers past conversations per user), conversation gating (3 free messages, sign in, 2 more, paywall, conversation preserved through every step), saved insights (personal reference library). Payment processing live via Stripe.
This week: Built 4 free tools powered by Colin's methodology. ICP Diagnostic, Cold Email Teardown, Pipeline Diagnosis, Outbound Week Planner. All free, no login, 30-60 seconds each. The thesis: free tools prove the mentor's value before money is asked. They rank on Google, they're shareable, and they funnel into the full agent.
Early signal: 21-minute average session in the first 48 hours. No ads, no growth hacks.
Revenue model: $47/mo platform + $150/mo per mentor agent. Mentors earn from every subscriber, and the agent extends their reach to every timezone and language without costing them hours.
Stack: Next.js, Supabase, Stripe, Claude API, Vercel.
What I'm figuring out now: distribution. Tools are the SEO play. But what else? Building in public here and looking for feedback.
Interesting concept. The real value will depend on how well the AI captures the mentor’s judgment, context, and decision-making process not just their knowledge.
If done well, it could make expert guidance far more accessible and scalable.
Curious how you ensure the AI stays aligned with the mentor’s evolving expertise.