Skip to content
Michi v2026.05.20
Save the Tokens

Bootstrap Your Project

Before your first planning session, run michi-bootstrap to set up the documentation structure Michi skills expect. This is a one-time setup — or re-run it later to close gaps.

Three phases, all interactive:

  1. Survey. The agent inventories your codebase — file count, language breakdown, existing docs, git history. It scans for Michi artifacts (CLAUDE.md, PROJECT.md, STATUS.md) and assesses what’s missing.

  2. Recommend. Based on the survey, it proposes an effort tier:

    • Greenfield (new or near-empty project): full scaffold — every Michi artifact created from scratch
    • Brownfield (existing project): gap analysis with S/M/L/XL effort tiers — you pick how much to set up now
  3. Execute. The agent drafts each document with you, not autonomously. It proposes content, you review, it adjusts. Nothing is generated and dumped.

DocumentPurposeAlways loaded?
CLAUDE.mdAgent rules, conventions, build/test commandsYes
PROJECT.mdGoals, users, features — the “why”Yes
STATUS.mdCurrent state, active work, what’s nextYes
ARCHITECTURE.mdDesign overview, component relationshipsOn demand
docs/ directoryEpics, plans, journals, verification

For brownfield projects, bootstrap assesses what you already have and only creates what’s missing. It won’t overwrite existing docs.

When I bootstrapped a Node.js MCP server project (existing code, no docs), the agent inventoried the codebase with tokei, scanned the file structure and git history, then proposed an L-tier setup — the full suite. It asked five questions before starting, I answered what was clear and deferred the rest, and it created seven artifacts conversationally.

One small correction along the way: it placed roadmap.md in docs/reference/ instead of docs/. I pointed it out, it moved it immediately. That’s the interactive nature — the agent drafts, you steer.

Update your CLAUDE.md with @-references pointing to your current work:

## Current Work
@./docs/epics/my-first-epic/plans/m1-plan.md

The @-reference mechanism controls what the agent loads at session start. Point it at the current milestone’s plan, and the agent has the context it needs.

Next: walk through your first milestone →