I’m building Archseed, and I’ve reached an uncomfortable point: some developers like the idea, a few have tried it, but nobody has bought the Pro plan.
Rather than post another feature announcement, I’d like to understand what I might be getting wrong.
The problem I’m trying to solve
When you give a coding agent an incomplete request, there are decisions it still needs to make:
What belongs in V1?
What should explicitly stay out?
Which architectural patterns should it follow?
Which assumptions need your approval before implementation?
My hypothesis is that making those decisions explicit before coding can help avoid unwanted scope and architectural choices.
For example, “build a booking app” leaves a lot unanswered. Does V1 need payments? Multiple businesses? Customer accounts? An agent can produce a plausible implementation without those decisions matching what you intended.
What Archseed does
Archseed is a CLI-first tool that guides you through scope, constraints, and architecture decisions, then generates structured project files you can use with your coding agent.
It’s not another coding agent, and it isn’t meant to replace your IDE.
An important limitation: generating Markdown does not automatically enforce architecture or guarantee that an agent will follow every decision. I don’t want to pretend otherwise.
Where I’m stuck
There’s a free workflow, and the current Pro beta is $12/month.
Some people have responded positively, but that hasn’t translated into paid users. Recent outreach also hasn’t led to many new trials.
I don’t yet know whether:
The problem isn’t painful enough.
The benefit isn’t clear enough to justify trying another tool.
Developers already solve it well enough with their agent’s planning workflow and project instructions.
A subscription is the wrong fit for something they mainly need when starting a project.
I don’t want to assume that lowering the price would fix the underlying issue.
What I’d like to learn
If you use coding agents, when was the last time one made a scope or architectural decision you had to undo? How did you handle it?
And looking at Archseed’s approach, what—if anything—would it add to your current workflow?
“I already handle this with my coding agent” is useful feedback too, especially if you can explain how.
You don’t need to install anything to reply. I’m more interested in understanding your experience than collecting signups.
Project: Archseed
Top comments (3)
I'd test the subscription part first. Planning is something people need at the start of a project and then forget about for weeks, so $12/month asks them to pay for the gap in between.
One thing you could try: sell a single scoped planning pass as a fixed offer. Something like "send me your app idea, you get the V1 scope, the out of scope list and the assumptions file in 48 hours, $49." Do part of it by hand at first. If people buy that, you see which decisions they actually care about, and the ones that come back every time tell you what the CLI should automate.
On your question, a scope call I see agents get wrong a lot is auth. They add accounts and roles to things that only needed a magic link. A short "not in V1" list in the project instructions fixes most of it, which is close to what you generate. So the value for me would be in the questions it forces me to answer, not in the Markdown it writes.
Thanks—this gives me something much more concrete to investigate.
Your distinction between the questions and the Markdown is especially helpful. The part I want Archseed to get right is surfacing decisions the developer hasn’t made yet, rather than just producing more documentation.
The auth example is a good illustration: whether V1 needs accounts, roles, or a simpler access flow is a decision to clarify before implementation.
I also take your point about frequency. I haven’t demonstrated a recurring reason to pay every month, so I shouldn’t assume the issue is simply whether $12 is affordable. A one-off, manually assisted planning pass sounds worth testing—while keeping in mind that selling that would validate a service first, not yet the CLI subscription.
When you write a “not in V1” list, how do you figure out what belongs on it? Is that usually obvious from the requirements, or do you have a set of questions you work through?
Respect for posting this instead of another feature announcement. My guess from my own side projects: devs liking the idea and devs paying $12/month are two different people. The planning step feels like something I can do in a markdown file myself, so the free tier already covers it in my head. I'd look for the moment where skipping planning cost someone real hours, and sell that moment instead of the workflow.