Meet Ferra. Your Claude Code setup, forged in 30 seconds.
Claude is only as useful as what you tell it about your project. Ferra handles that part — she reads your repo, matches your stack against 59 curated skills, and generates a production-ready CLAUDE.md in 30 seconds.
Built for developers who'd rather ship than configure.
Claude is only as good as what you tell it about you.
We kept opening new projects and doing the same thing — writing CLAUDE.md from scratch, copy-pasting snippets, wondering if we missed a skill. Every time. Forge exists so you never have to do that again.
A developer who spends 30 minutes configuring Claude properly will outcode someone who skips it entirely — every single session, indefinitely.
01
Setup friction kills adoption
Most developers never configure CLAUDE.md. Not because they don't care — because figuring out what to put in it takes longer than the first feature they want to ship.
02
Generic prompts get generic code
Claude doesn't know you use Prisma, not raw SQL. Or that you're on App Router, not Pages. Without that context, it guesses — and guesses wrong half the time.
03
Every stack deserves a custom context
A Go service and a Next.js SaaS shouldn't share the same CLAUDE.md. Forge detects your actual stack and selects the skills that matter for it.
How it works
From repo to ready in 3 steps.
01
Tell Ferra about your project
Paste a GitHub URL — Ferra runs a config-first scan. Or describe your stack in plain text.
02
Ferra selects your skills
She detects your stack and picks the right skills from her library of 59. Your CLAUDE.md is generated with Karpathy-inspired principles.
03
Drop it in your repo
Download a ZIP with CLAUDE.md, skill files, hooks, and a setup guide. Claude Code auto-reads it every session.