Forest: From idea to launched product, without building on a guess
For founders and designers building their own product with AI. Forest runs inside Claude Code. It asks why before it asks what, and records how you know each answer.
How it works in 3 steps
Tell it the idea
It asks who it's for and what you want at the end. A vague answer gets a follow-up question.
Check it, then plan it
You talk to real people. It weighs what you heard, then writes the requirements and a design direction from your references.
Build and launch
Claude Code builds it phase by phase, checks each phase against the design, and ships it.
Slower, on purpose
Some tools write a PRD in 90 seconds, on guesses. Forest asks first, and tags every answer by how you know it.
Watched
You saw someone do it.
Heard
A named person told you, unprompted.
Reasoned
Worked out from what you saw or heard.
Assumed
Nobody has checked yet.
Would it let this through?
Moments from real projects, dealt fresh each visit. Call each one, then turn the card.
01 / 08
You say it's for “small businesses”.
Would forest let this through?
Not quite.
It pushes back.
It asks for one person you have actually spoken to.
Card 1 of 8.
What you leave with
Each camp works from what the one before it made. Stop at any flag and you leave with what is there.
Camp 1
Frame.
- Needs
- Your idea, however rough. No idea yet? It runs an idea hunt first.
- Makes
- Who it's for, by name, the job they're trying to do, and the belief that sinks it if wrong.
who + the betLearnCamp 2
ProofLearn.
- Needs
- A named person and the riskiest belief, from Frame.
- Makes
- Interviews and a stress test, then
EVIDENCE.md: every claim tagged by how you know it.
EVIDENCE.mdDecideCamp 3
PlanDecide.
- Needs
- The gate's reading, press on, reroute or turn back, and the evidence behind it.
- Makes
- A PRD: requirements someone else can check, and scope in, out and later, with something cut.
PRDShapeCamp 4
PrototypeShape.
- Needs
- The PRD's scope, and references you like.
- Makes
design.md: colour, type, spacing and voice, each traced to a reference, every contrast checked.
design.mdBuildCamp 5
Build.
- Needs
- The PRD and
design.md, compiled intoBRIEF.md. - Makes
- The product, phase by phase, each one reviewed, tested and compared with the design.
tested buildShipCamp 6
LaunchShip.
- Needs
- A build with no open blockers.
- Makes
- The live product,
SUMMIT.mdon what actually went live, and how it reaches people.
Stop here and you leave with the product, live.
Every document, on one page
Open project/index.html, or ask /map to open it. It stays on your computer.
Every document, as written
Each file appears under its camp, with the command that wrote it and when. The ones still owed say so.
Rebuilt as you work
Every write rebuilds it, and an open tab refreshes itself. It holds nothing the folder doesn't.
Copy what you need
A whole document or one section, as written or as a prompt for another AI tool.
Questions people ask
Why forest kit?
Because an AI tool will write you a convincing plan in seconds, and nothing in it tells you which parts are true. Forest asks before it writes, tags every claim by how you know it, and can tell you to stop before you build. Each decision is kept with its reason, so the folder still explains itself 2 to 3 months on, to you, a developer or an investor. It's slower than working freehand. Going fast on a guess gets you nowhere sooner. That's the trade.
What are camps?
Forest splits the work into 6 stages it calls camps: Frame, Learn, Decide, Shape, Build and Ship. Each one ends with files you keep.
How long does it take?
The first session takes 20 to 40 minutes. After that, it depends on how long your research takes.
Can you stop and come back?
Yes. Everything lives in the folder. Open it again and type /trail.
What happens after launch?
SUMMIT.md leaves 3 questions dated 90 days out. Answer them cold, and /pack opens cycle 2 with your research, decisions and design carried across.
Does it send your work anywhere?
No. It checks for a new version at most once a day. Usage counts go out only if you say yes, and never your files.
Why it exists
For founders
Plenty of conviction, no evidence yet, and no way of knowing whether the idea solves the right problem.
For design founders
Endless exploration: round after round of ideas, comments on comments, documents nobody reads, and a brief that still isn’t clear.
For whoever comes next
A Slack channel full of ideas, AI‑written documents and decisions nobody checked, and a CMS about to become the bottleneck.