Our first customer refused to use our website. So we shipped MCP access instead.
When we launched, the product was simple: chat with an AI agent trained on a real sales expert's methodology. Web-based, conversational.
Our first paying customer changed that. He didn't want to chat on our site. He wanted to plug the expert into his own agent stack. His agent handles customer interactions, and when it hits a sales question, it queries our expert module instead of guessing.
His words: "I just add it and boom, my agent has the sales stuff."
So we built it. Published an MCP server on npm (`@forgehouseio/mcp-server`). Two tools: list_mentors and ask_mentor. Add it to Claude Desktop, Cursor, or any MCP-compatible agent.
The interesting part: he chose the agent module over scheduling a call with the actual human expert. "I'm not ready yet," he said. The module removes the pressure of a human interaction while delivering the same frameworks.
We're now thinking about this as agentic expertise modules. Not mentoring. Not coaching. Vetted human knowledge, structured and queryable by agents.
Now reworking our mentor acquisition strategy and ICP based on this insight.
it's been similar to us - and pattern seems to be popping up across industries. It's not anymore future where human-computer interface is used, but rather agent-computer. At least, it'll be just as same big market as human-computer interface.
right, and it seems like we have to meet the user where he is at and make our product/solution accessible via the agent interface they are using already.
This is exactly what I'm seeing too. I'm building a desktop app for saving/reusing UI components across Figma, Webflow, etc. — and the most useful external interface turned out to be an MCP server, not a REST API or a browser extension.
The pattern is the same as yours: users don't want to leave their workflow. A designer in Cursor or Claude can query their component library without switching apps. The tool becomes invisible infrastructure.
Your insight about the customer preferring the agent module over scheduling a human call is fascinating. Removes the social friction entirely. Curious how you're handling the "trust gap" — does the customer trust the agent's answers as much as a live call?
That's a clean use case. UI components as MCP tools makes a lot of sense, the designer never leaves their editor.
On the trust gap: our first customer actually said he's "not ready yet" to talk to the actual human expert. He chose the module first. So in his case, trust wasn't the barrier to the agent, it was the barrier to the human. The module let him engage with the expertise without the social pressure of a live call.
Whether he trusts the agent's answers as much as the human's, honestly I don't know yet. We have one paying customer. But the signal so far is that for this type of user (technical founder, builds with agents), the convenience and low friction of the module outweighs whatever trust premium a live call would have.
Thanks for sharing that insight about the trust gap going the other way — the user trusting the module before the human. That's a really interesting inversion.
For my case it's similar in a way: designers don't want to leave their editor to use a separate tool. The clipboard is the lowest-friction interface possible — they already copy-paste constantly. The MCP layer just makes it programmable.
One paying customer is a real signal though. Curious how the usage patterns evolve as you get more.
Hello Indie Hackers! 👋
I'm excited to share that my latest micro-SaaS, SachCheck AI, just got approved and featured on the SideProjectors homepage!
The Problem:
In India, fake news in regional languages like Hindi spreads like wildfire. Most tools are built for English, leaving 600M+ Hindi speakers vulnerable.
The Solution:
SachCheck AI is a lightweight tool that uses the Google Fact Check API to verify claims instantly in Hindi.
Tech Stack:
- Frontend: Vanilla JS, HTML, CSS
- Hosting: Vercel
- API: Google Fact Check Tools API
I am now looking for a new owner to take this forward and scale it. You can see the live listing here: https://www.sideprojectors.com/project/sach-check-
Would love your feedback on the tool!