You type one line. It reads your repo, asks the few questions that matter, plans the work in front of four reviewer agents, parallelises into worktrees, and ships a PR. Below, what each step actually does.
One curl. The signed binary lands in ~/.local/bin and registers itself with whichever AI IDE you already use. Claude Code, Cursor, OpenCode, Continue, Windsurf, aider. macOS, Linux, Windows. No config files to write.
$ curl -fsSL https://creativehuman.ai/brew-chai/install.sh | bash → Detecting platform... darwin/arm64 → Verifying Ed25519 signature... OK → Installing brew_chai 1.0.0 to ~/.local/bin ✓ Installed · 5 platforms supported
/brew_chai:brew "<what you want>". The pipeline classifies intent (build / fix / heal / tool), reads your repo conventions, and starts. You don't pick a model. You don't craft a prompt.
$ /brew_chai:brew "add OAuth login to the settings page" [1/8] setup ✓ branch from main · conventions loaded [2/8] scope ✓ build · feature-add · scope=medium
Phase 03 surfaces only what's genuinely uncertain. High-stakes questions (affects tests, UI, security) get asked. Low-stakes questions (formatting, naming) get auto-answered with sensible defaults. One round. No follow-up.
[3/8] ambiguity found 4 questions · 1 high-stakes [4/8] q&a ? Which providers? [high-stakes: affects tests + UI] › google, github · 3 low-stakes questions defaulted
Phase 05: a draft plan goes to four reviewer agents. They critique. The plan revises. The cycle continues until a numeric gate (>95/100). If it can't get there, you get an escalation log explaining why.
[5/8] plan → draft · v1 · score 89/100 → critique · 4 reviewers · 12 findings → revise · v2 · score 94/100 → critique · 4 reviewers · 3 findings → revise · v3 · score 97/100 ✓ plan accepted
Phase 07 decomposes into sprints, then runs them in parallel git worktrees. An audit-fix loop runs per task. The coverage gate (100/100/100/100 on changed lines) is non-negotiable. Failed tasks retry with the auditor's notes.
[6/8] decompose ✓ 3 sprints · 11 tasks · graph clean [7/8] execute → parallel agent swarm in worktrees ✓ sprint 01 · provider scaffold (cov 100/100/100/100) ✓ sprint 02 · settings UI integration (cov 100/100/100/100) ✓ sprint 03 · session middleware (cov 100/100/100/100) ✓ deep-clean · 6 auditors · all green
Phase 08 ships: 11 commits on average, a PR description that reads the diff, an escalation log if anything blocked, and extracted knowledge that feeds future runs. Ready to merge.
[8/8] handoff ✓ PR #1284 · 11 commits · 0 deferrals → DONE · 47 min · 0 mid-run interrupts PR description written from diff Escalation log empty ✓ Knowledge extracted to .brew_chai/knowledge.md
Below: an actual brew of "add OAuth login to the settings page" running through all eight phases. 47 minutes wall-time. Zero mid-run interrupts after the Q&A.