Requires Node.js ≥ 18.
Option A — npm (recommended)
npm install -g @episuarez/init-claudeOption B — git clone (with auto-update)
git clone https://github.com/episuarez/initialiser %LOCALAPPDATA%\init-claude
cd %LOCALAPPDATA%\init-claude && npm install --omit=devAdd %LOCALAPPDATA%\init-claude to your PATH.
init-claude update will pull new versions automatically.
Then, from any project:
init-claudeinit-claude updateThis pulls the latest version (git pull + npm install) and never breaks again:
- If your install is a git clone (Option B, or
install.cmdon a machine with Git), it just updates. - If your install is an old copy-based one (xcopy, no
.git),updateauto-converts it to a git repo in place the first time, then updates — no manual reclone. - npm installs (Option A) update with
npm update -g @episuarez/init-claude.
If your install predates this auto-migration and
updatestill complains it's not a git repo, re-runinstall.cmd(now clones via Git) or switch to npm. After that,init-claude updateis all you ever need.
Scans your project (languages, frameworks, size, CI, docs, design files), recommends the right Claude Code components and skills, and installs them — MCPs registered, CLAUDE.md written, git hooks wired.
| Component | Group | What it does |
|---|---|---|
| context-mode | Core | Session memory + output sandbox. Cuts context waste by ~98%. |
| sequential-thinking | Core | Structured reasoning for architecture and design decisions. |
| context7 | Core | Up-to-date library docs (React, Next.js, FastAPI…) fetched at runtime. |
| RTK | Core | Compresses Bash output 60–90% (git, npm, cargo, test runners). |
| autoskills | Skills | Auto-installs curated, SHA-256-verified skills matched to your stack (React, Next, Vue, Zod, Tailwind…) via npx autoskills. |
| code-review-graph | Analysis | Codebase dependency graph: impact analysis, multi-file search. |
| serena | Analysis | Semantic LSP: go-to-definition, find-references, precise symbol search. |
| playwright | Web | Browser automation, scraping, E2E testing via MCP. |
| claude-flow | Orchestration | Multi-agent swarm / hive-mind for large parallel tasks. |
| SuperClaude | Orchestration | 30 /sc:* workflow slash commands. |
| husky + lint-staged | Git | Pre-commit hooks: lint and tests before every commit. |
| MCP Gateway | Optimization | Consolidates many MCP servers behind one endpoint (~50% fewer tokens with a large MCP fleet). Opt-in. |
| MarkItDown | Documents | Converts PDF/Word/Excel/PPT to Markdown for Claude to read. |
| Pencil | Design | Vector design .pen files + pencil-to-code skill. |
| Figma Dev Mode | Design | Reads Figma designs (requires Figma desktop + Dev seat). |
How defaults are chosen — by cost asymmetry, not project size: zero-overhead universal wins (context-mode, sequential-thinking, context7, RTK) are always on; heavy MCPs (serena ~20 tools, playwright ~25, claude-flow +87) are pre-checked only with a real applicability signal — never just because the repo is big. claude-flow and Figma are opt-in. Anything already provided by an installed plugin or MCP is skipped to avoid duplicate tools.
Project skills installed per-stack: frontend-components · api-design · unity-conventions · e2e-testing · python-quality — and, via the autoskills component, any stack-matched skills from the autoskills registry.
init-claude Interactive wizard (recommended)
init-claude --yes Apply recommended defaults without prompts
init-claude check Doctor: installed tools, MCP Tool Search state, CLAUDE.md size
init-claude update Self-update the app (git pull + npm install)
init-claude upgrade Update installed components (npm, pipx, cargo)
init-claude add-skill <id> Install a skill from the built-in catalog
init-claude add-skill <url> Install a skill from any raw URL or GitHub repo
init-claude --version Print version and exit
On first run, init-claude creates user-rules.md in the app directory. Anything you write below the --- separator is injected as a section into every CLAUDE.md the tool generates, in every project.
- Gitignored —
init-claude updatenever touches it. - Edit once, every future run picks it up automatically.
- Per-project rules go in the
CUSTOMblock of each project'sCLAUDE.md(also survives regeneration).
%LOCALAPPDATA%\init-claude\user-rules.md ← global personal rules (all projects)
<project>\CLAUDE.md CUSTOM block ← rules for that project only
Context-mode stores data globally — not in your project:
~/.claude/context-mode/
├── content/ indexed tool outputs (SQLite, hashed names)
└── sessions/ session memory (same scheme)
Query from inside Claude Code with ctx_search or ctx stats. Don't browse the files directly.
Add a component — edit catalog/components.json. No code changes needed:
Add a project skill — drop a .md file in catalog/skills/ and add a record to projectSkills in components.json.
Remote skill registry — create config.json with { "registryUrl": "https://..." } pointing to a { "skills": [...] } JSON. The wizard will offer those skills on every run.
See CONTRIBUTING.md for full details.
Option A — npm installs globally via npm. Update with npm update -g @episuarez/init-claude. No git required.
Option B — git clone supports init-claude update (git pull + npm install) and lets you edit the catalog locally. Checks for updates once a day silently.
Option C — install.cmd — clones via Git when available (auto-updatable with init-claude update) and updates PATH. Without Git it falls back to a plain copy (no auto-update). Re-running it on an old copy-based install reconverts it to a git clone.
init-claude upgrade updates the installed components (Claude Code CLI, context-mode, pipx packages, RTK) regardless of which installation option you used.
- Node.js ≥ 18
- Git (option A + auto-update)
- Claude Code CLI —
npm install -g @anthropic-ai/claude-code - Python + pipx — optional (code-review-graph, MarkItDown, SuperClaude)
- Rust + cargo — optional (RTK)
- uv — optional (serena)
MIT with Commons Clause. Free for personal, open-source, and internal use. Attribution required on forks. Commercial redistribution requires a separate license — contact.
See LICENSE for full terms.
Epifanio Suárez Martínez — if you fork this, keep the attribution. If you build something commercial on top of it, let's talk.

{ "id": "my-tool", "name": "My Tool", "group": "Analysis", "desc": "Short description shown in the wizard.", "recommendIf": ["frontend", "large"], "alwaysOn": false, "install": { "type": "npm", "pkg": "my-tool" }, "mcp": { "name": "my-tool", "cmd": "npx -y my-tool-mcp" }, "claudemd": "- `my-tool`: one-line hint injected into CLAUDE.md." }