AI agent tools charging $200/month are a scam. Here's proof.

Hard agree on the pricing thing - I looked at a few of these tools and the "agent" part was just LangChain + a couple API wrappers. Now I just build the integrations myself and it's literally 10x cheaper.

you're not paying for the capability. you're paying for someone else to make the decisions. fair trade for some, but terrible deal for builders.

The ‘pay for convenience’ model only works until someone like you ships the open source version. Respect the transparency on the catch too most builders would just hide that.

The Buyer A / Buyer B framing is useful, but I think there's a third category that's becoming more relevant by the month: the semi-DIY founder.

I work primarily with solopreneurs who sit in the middle — they're technical enough to use APIs and configure tools, but they don't want to build and maintain a full agent stack. For them, the real unlock hasn't been either extreme (build from scratch or pay $200/mo). It's been curated prompt patterns and workflow templates that sit on top of the APIs they already have access to.

The interesting shift I'm seeing: as models get better at following nuanced instructions, the "moat" of a tool like Manus isn't the agent architecture — it's the prompt engineering and workflow design that makes it useful for specific tasks. And that's something that can be packaged and shared without needing a whole SaaS subscription.

So there's a viable middle path: buy the API credits ($5-15/mo as you said), use well-designed prompt patterns for your specific use cases, and skip both the $200/mo tool AND the multi-day build effort. The semi-DIY approach is how most of the founders I know are actually operating right now — they just don't talk about it because it's less sexy than either extreme.

The pricing argument is fair. But there's a hidden cost people don't count: their own time.

I've built my own agent setups. Every. Single. Time. I spend 2-3 days getting it right, then a week later something breaks (API changes, dependency conflicts, whatever) and I'm back in maintenance mode.

For a solo founder, $200/month is a lot. But $5-15/month + 10 hours/month of maintenance? That's way more expensive if you value your time at all.

The real question isn't 'can I build it cheaper?' It's 'can I build it and forget about it?' Most of the time the answer is no.

The pricing argument is interesting.

What I'd be most curious about is whether people are actually paying for the capabilities, or paying to avoid becoming the person responsible for making all the infrastructure decisions themselves.

Those can look like the same purchase today, but they tend to lead to very different conclusions about who the real customer is.