Research your market
Agents study your market, competitors, and tools before you write a line of code.
Your agents do the homework
Most product discovery is a blank Google Doc and three hours of staring at it. You try to think through pricing, competitors, features, and technical decisions all at once, and end up with a messy document that doesn't actually help when it's time to build.
This flips it around. You describe your idea, and agents go research it for you. They come back with competitor analysis, tool recommendations, pricing intelligence, and demand validation. By the time they ask you product questions, they already know the market.
How it works
Open Claude Code and type /discover. Then just describe what you want to build in plain English. A reference URL helps (a competitor, a similar product, something you want to be like) but it's not required.
The system runs in phases. First it makes sure it understands your idea well enough to research accurately. Then three research agents fan out in parallel:
One studies the market. Competitors, pricing patterns, what users love, what they complain about, gaps nobody's filling. Another finds the external tools and APIs your idea needs, compares options, and picks the best fit for each. A third validates demand by looking at search trends, social signals, and whether similar products have traction.
You review what they found. If something's off, they adjust. If it looks right, the system moves into product discovery. Five agents with different perspectives (product, experience, brand, architecture, market analysis) work through a structured conversation with you. They already have the research, so they're not asking questions they could answer themselves. They propose answers and ask you to confirm or adjust.
What comes out
Seven documents that describe your product completely:
- What the product is, who it's for, and how you'll make money
- Your users, their journeys through the product, onboarding flow, and conversion funnels
- Every feature, prioritized into MVP, V2, and future
- Brand personality, colors, typography, and voice
- Technical architecture, data model, and security approach
- External tools beyond the standard stack
- UX patterns for errors, loading states, empty states, and destructive actions
Every downstream command reads these docs. The specs reference them. The agents reference them. The design system pulls from the brand guidelines. The billing workflow reads the pricing. Nothing gets built in a vacuum.
What this replaces
Weeks of product thinking compressed into one session. You still make every important decision, but the grunt work of researching competitors, comparing tools, and structuring documentation is handled for you. The output is specific enough that agents can build from it without coming back to ask "what did you mean by that?"