Took me 18 months but I finally shipped. AI agents that actually capture subscribers and make money.
Finally shipped the thing I've been working on forever.
Quick backstory: I kept building AI agents that people would use once, get their answer, and bounce. No email. No payment. Nothing.
The AI part wasn't the problem. That's easy now. The problem was everything around it.
The stuff that was annoying me:
- People use my agent → leave → I have no way to reach them
- I want to charge for access but there's no easy way
- I want it on Discord AND my website AND Slack but that's like 3 different builds
- Connecting to Calendly or my email list required hacky workarounds
Couldn't find anything that just... did all of that.
So I built it.
What it does:
Upload your content. Get an AI agent. Then:
- Gate it behind email (they sign up before chatting)
- Or charge for it (subscription or pay-per-chat)
- Connect to 50+ tools (Calendly, Stripe, Mailchimp, HubSpot, etc.)
- Deploy anywhere (web, custom domain, Discord, Slack, Telegram, WhatsApp)
The cool part — interactive responses:
The agent doesn't just spit out text. It can actually do things.
- "Design a thumbnail" → opens canvas editor → user edits → exports PNG
- "Build a landing page" → generates React + Tailwind → live preview → user edits → exports code
- "Analyze this data" → writes Python → runs it → returns charts
- "Book a call" → shows Calendly widget → books directly
It's the difference between "here's how to do it" and "here, it's done, tweak it if you want."
What people are building with it:
- AI tutor that charges $29/mo
- Design assistant that captures emails and upsells to paid
- Code reviewer that charges per review
- Customer support bot that actually handles forms and bookings
Being honest about the timeline:
- V1 has been live since July 2023
- Took me 18 months to rebuild V2 (yeah, way too long)
- Just shipped today
- Free tier is real, not a trial
What I learned:
- Building in silence sucks. Should've shipped updates publicly.
- The AI isn't the product. The stuff around it is.
- People care more about "can it book calls?" than "14 artifact types"
Would love feedback:
- Does this solve something for you?
- What would you build?
- What's missing?
Link: https://boostgpt.co
Thanks for reading ✌️
Trending on Indie Hackers
"The AI isn't the product. The stuff around it is." — this is the single most important insight in AI product building right now and most people still don't get it.
I've been building some browser-based dev tools and keep running into the same pattern: the "smart" part takes 10% of the effort, the capture/monetization/distribution layer takes 90%. You can have the best model in the world, but if users can't easily share it, pay for it, or integrate it into their workflow, it's just a tech demo.
The interactive responses angle is clever. Turning chat from "here's an answer" into "here's a working thing you can edit" changes the value prop entirely. That's the difference between a chatbot and a product.
18 months is a long time, but the fact that you rebuilt V2 from scratch based on real feedback ("I just want to brand it and share it, no code") shows you were actually listening. Most people would've just kept adding features to the broken foundation.
Congrats on shipping. The "building in silence" regret is real — curious how the public shipping phase goes.
Yeah the 10% / 90% split is real. I spent way too long on the AI part early on, thinking that was the product. It's not. It's table stakes now.
The browser dev tools thing sounds familiar. What are you building?
Curious if you're hitting the same "users want it simpler than I made it" problem I had.
On the interactive responses... that was actually the unlock for V2. V1 was basically "chat and get text back." Useful but forgettable. Once the agent could actually do stuff... show a canvas, run code, display forms... it stopped feeling like a chatbot.
People started using it differently. More like a tool than a conversation.
The rebuild decision was hard. V1 had users.
Throwing away a working thing to start over feels wrong. But every new feature I tried to add was fighting the foundation. At some point you have to admit the architecture doesn't fit what you're actually building.
Public shipping is going well so far. 5 weeks in, 3 products out. Had a delay this week (PageRemix slipped by 2 days) but I just said what happened and kept going. That's the part I like about building in public... no hiding. You either ship or you explain why you didn't.
I appreciate the kind words. What's the timeline on your dev tools? Would be curious to see what you're working on.
I can relate to this in a good way. ~
The term "AI agents" does not do justice to what you shipped, which closes the loop around intent. The pattern you have described – agents get used by people once, get value, and disappear – is where most AI demos die. You did not repair the model, rather you repaired everything that occurs before and after the answer.
I observed a few things that stood out to me. The AI is not the product. The items that surround it
I have also seen this too. Users naturally do not wake up in the morning wanting intelligence; they want outcomes. The actual products are email capture, payments, booking, and follow ups.
Responses that are interactive are greater than those that are static text.
Once an agent is capable of performing any action, such as booking a reservation, developing a program, or executing a code, it appears to change. It feels no longer like a toy and starts feeling like software.
That is a huge mental change.
It takes 18 months to recover a deeply relatable V2.
It is especially true when you only realize late that distribution + monetization primitives should’ve been first-class citizens from day one.
It might be helpful to use this frame moving forward.
It seems as though you are constructing a monetization layer for expertise, not an agent builder. That could be a more resonant wedge than “50+ integrations” or “artifact types.” 👏
You're right... Nobody cares about 50+ integrations. They care about "can I make money from what I know?" That's a tighter story. Course creators, coaches, consultants... they want to package their expertise and get paid. The AI is just the delivery mechanism.
The 18 months still stings. I built the interesting technical stuff first, left monetization for later. Should've been the opposite.
Appreciate this. 🙏
Your product looks great, so does your journey.
Thanks Joey 🙏
What a ride! Interesting journey and great learning lessons 👌
Thanks! Yeah it's been a long road but finally out there. Appreciate you reading through it 🙏
18 months is a long time but sometimes that's what it takes to build something solid. The insight about "the AI isn't the product, the stuff around it is" really resonates - seeing this pattern a lot in successful AI products.
The multi-channel deployment is smart. Most founders I've talked to (including some interviewed on the Public SaaS Builders podcast) struggle with the same thing: great AI, terrible distribution. You're solving the right problem.
Question: what made you decide to rebuild V2 from scratch vs iterating on V1? That's always a tough call for solo founders.
Thanks! Yeah the "AI isn't the product" thing took me a while to learn. I kept building cool AI features and wondering why people weren't excited. Turns out they just wanted to set it up, connect their tools, and share a link. The AI part they assumed would just work.
On the rebuild vs iterate question: honestly it wasn't an easy call.
V1 was very API-focused. Powerful if you were technical, but most people bounced.
The feedback I kept getting was: "I just want to brand it, connect my stuff, and share it. No code."
I tried patching V1 at first. Added some UI, made things easier. But the architecture was built around "developer builds on top of our API" — not "creator sets up agent in 5 minutes."
Every new feature felt like I was fighting the foundation.
So I made the call to rebuild. Took way longer than I wanted (18 months... yeah). But V2 is actually built for the use case people were asking for. Branded agent pages, no-code connectors, interactive artifacts, email agents... none of that would've fit cleanly into V1.
Looking back, I think it was the right call. But I'd be lying if I said there weren't moments where I regretted it.
Would love to check out the podcast btw... got a link?
Congrats on shipping after 18 months. That's a long grind but the result looks solid.
Your framing nails it: "The AI isn't the product. The stuff around it is." This is what most AI-wrapper builders miss. The AI part is commoditized — the capture, monetization, and distribution layer is where the real value sits.
The interactive responses concept is interesting. Embedding actual tools (canvas, code execution, booking) inside the chat changes the value prop from "assistant" to "automated workflow." That's a meaningful upgrade.
A few questions:
Email gating vs. friction — Do you see drop-off when people hit the email wall before chatting? Or does pre-qualifying intent actually improve downstream conversion?
Multi-platform deployment — When someone deploys the same agent to Discord + web + Slack, how do you handle context persistence across channels? (Does the agent "remember" a user across platforms?)
What surprised you most about user behavior in V1 that shaped V2?
The "building in silence sucks" lesson is real. Curious how you plan to approach distribution this time — IH, Twitter, somewhere else?
Thanks! Yeah, 18 months was way too long. Lesson learned on that one.
To your questions:
Email gating vs friction
Honestly, I expected more drop-off than I saw. Turns out if someone actually wants to use the agent, they'll give their email. The people who bounce at the gate probably weren't going to convert anyway. It's like a natural filter.
Multi-platform context
Right now each platform is separate. So if someone chats on Discord then goes to web, the agent doesn't know it's the same person. It's on my list to fix... probably tie it to email if they've logged in on both. But honestly most users stick to one channel, so it hasn't been a huge pain point yet.
What surprised me from V1
V1 was very API-focused. You could build cool stuff with it, but it required technical know-how to get the most out of it.
The feedback I kept hearing: "I just want to set up my agent, brand it, connect my tools, and share it. I don't want to write code."
That shaped everything in V2:
Basically went from "here's an API, figure it out" to "here's your agent, it's ready to go."
Distribution this time
Mainly Twitter and Youtube for now. Came back to Twitter after 18 months of silence, been shipping in public for the last few weeks. No big Product Hunt launch yet... want to get more feedback and testimonials first, then do PH when I have some social proof.
What about you... building something in this space?
Thanks for the detailed answers.
Email gating as natural filter — that framing makes sense. The friction isn't a bug, it's a feature. Filters out tire-kickers early.
Multi-platform context — tying it to email is probably the cleanest solution. The "most users stick to one channel" insight is useful too. Solving for edge cases that rarely happen is a common trap.
V1 → V2 shift — "API-first to no-code" is a pattern I keep seeing. Developers build for developers, then realize the market is actually creators who just want things to work. Your pivot sounds like it was painful but necessary.
To your question: yes, building in a related space. I'm working on Stream Tech — a tech news aggregator with AI-powered summaries. Different angle (content curation vs agent building), but similar underlying challenge: making AI useful without requiring users to understand how it works.
The "stuff around the AI" lesson applies here too. The AI summarization is the easy part. The hard part is curation, personalization, and delivery (Slack integration, RSS, digest emails).
Good luck with the Twitter/YouTube distribution. Shipping in public after 18 months of silence takes guts. Following your progress.
congrats on finally shipping! 18 months is a long time to stay in the cave but sounds like you came out with something solid.
the "building in silence sucks" lesson hits hard. im going through that now with my own project and the isolation is real. forcing myself to post updates even when it feels premature.
question on the email gating - do you give people any preview of what the agent can do before they hand over their email? or is it more like "trust me, sign up first"? curious how you balanced that.
Thanks! Yeah 18 months was way too long. I don't recommend it.
On the isolation thing... I feel you. The urge to wait until it's "ready" is strong. But ready never comes. Just post the ugly stuff. Nobody judges as harshly as you think they will.
For the email gating question: you can configure it either way actually.
some people do full gate: you hit the agent page, you see the branding and description, but you have to sign up before you can chat at all.
others do soft gate: let people send a few messages first, then ask for subscription to continue. kind of a "try before you commit" thing.
I've seen both work. full gate gets fewer but more serious leads. soft gate gets more signups but some are just curious and never come back.
Depends on what you're optimizing for... list size or lead quality.
What are you building?
Happy to share more if it's relevant to your use case.
This comment was deleted 2 months ago.