Book a call
07 / Case study · 2025

asBuilt

Two documentations from one codebase.

  • RoleProduct + AI architecture
  • KindAI codebase documentation generator
  • Repo Lezzur/asBuilt ↗
asBuilt — AI codebase documentation generator
Inside the system
  • asBuilt — Docs Docs
  • asBuilt — New Scan New Scan
  • asBuilt — Results Results
  • asBuilt — Scan Complete Scan Complete
  • asBuilt — Scanning Scanning

Click any image to view full size.

Feed asBuilt a zip, a local folder, or a GitHub repo. It analyzes the codebase and emits two documents: a technical reference tuned for AI coding assistants, and a human-readable overview for stakeholders. Multi-provider LLM support.

By the numbers
  1. Zip / folder / GitHub Input sources
  2. 2 (AI + human) Outputs per run
  3. Multi-LLM Providers supported
Problem

Codebase documentation is usually either too technical for stakeholders or too hand-wavy for coding agents. Teams end up maintaining two versions — which means neither stays up to date.

What I built

asBuilt ingests a codebase, walks the file tree, summarizes at the file and module level, and produces two outputs from the same pass: a dense AI-readable reference (with module boundaries, call graphs, invariants) and a human overview (with the why, not just the how).

Outcome

Stakeholders read one document, coding agents consume the other, both are regenerable on commit. Documentation stops being a second job.

Stack
  • Next.js
  • Firebase
  • Multi-LLM routing