dotai scans your codebase and auto-generates configuration files for every major AI coding assistant — Claude Code, Cursor, Windsurf, GitHub Copilot, and OpenAI Codex. No API keys. No accounts. Works completely offline.

npx dotai

That's it. Five config files. One command.

The Problem

Every AI coding tool has its own config format. Setting up a new project means:

# Without dotai — manual, repetitive, error-prone

1. Read Claude Code docs → write CLAUDE.md by hand
2. Read Cursor docs → write .cursorrules by hand
3. Read Windsurf docs → write .windsurfrules by hand
4. Read Copilot docs → write .github/copilot-instructions.md by hand
5. Read Codex docs → write codex-instructions.md by hand
6. Repeat for every new project
7. Keep all 5 files in sync when your stack changes

# With dotai — automatic, consistent, always accurate

npx dotai    # done

Your AI tools go from "generic JavaScript suggestions" to "knows you use Next.js 14 with Prisma, pnpm, Vitest, and deploy to Vercel."

Quick Start

# Run in any project directory
npx dotai

# Preview without writing files
npx dotai --dry-run

# Generate only Claude and Cursor configs
npx dotai --format claude,cursor

# Guided interactive setup
npx dotai --interactive

Example Output

📦 my-app — Codebase Analysis

  Language:         TypeScript (85%), JavaScript (15%)
  Framework:        Next.js 14.1.0
  Package Manager:  pnpm
  Testing:          Vitest
  Linter:           ESLint
  Formatter:        Prettier
  Database:         PostgreSQL, Redis
  ORM:              Prisma
  Deployment:       Docker, Vercel
  CI/CD:            GitHub Actions
  Monorepo:         Turborepo (3 packages)
  API Style:        REST, tRPC

✨ Generated 5 files:
  ✓ CLAUDE.md
  ✓ .cursorrules
  ✓ .windsurfrules
  ✓ .github/copilot-instructions.md
  ✓ codex-instructions.md

Before / After

Before — no AI config

my-app/
├── src/
├── tests/
├── package.json
├── tsconfig.json
└── .eslintrc.json

Every AI tool gives generic, context-free suggestions. Copilot doesn't know you use Prisma. Claude doesn't know you prefer Vitest. Cursor doesn't know your monorepo layout.

After — npx dotai

my-app/
├── src/
├── tests/
├── package.json
├── tsconfig.json
├── .eslintrc.json
├── CLAUDE.md                        ← generated
├── .cursorrules                     ← generated
├── .windsurfrules                   ← generated
├── .github/copilot-instructions.md  ← generated
└── codex-instructions.md            ← generated

Every AI tool now knows your exact stack, conventions, commands, and architecture.

Output Formats

| File | AI Tool | Description | |------|---------|-------------| | CLAUDE.md | Claude Code | Project instructions for Anthropic's CLI agent | | .cursorrules | Cursor | Rules file for Cursor IDE's AI assistant | | .windsurfrules | Windsurf | Rules file for Windsurf (Codeium) IDE | | .github/copilot-instructions.md | GitHub Copilot | Custom instructions for Copilot | | codex-instructions.md | OpenAI Codex | Instructions for the Codex CLI tool |

Detected Technologies

Languages

TypeScript, JavaScript, Python, Rust, Go, Ruby, Java, PHP, C#

Package Managers

| Manager | Detected via | |---------|-------------| | npm | package-lock.json | | yarn | yarn.lock | | pnpm | pnpm-lock.yaml | | bun | bun.lockb / bun.lock | | cargo | Cargo.toml | | go | go.mod | | pip | requirements.txt / pyproject.toml | | uv | uv.lock | | poetry | poetry.lock | | bundler | Gemfile.lock | | composer | composer.lock | | maven | pom.xml | | gradle | build.gradle / build.gradle.kts | | dotnet | *.csproj / *.sln |

Frameworks

| Category | Frameworks | |----------|------------| | React ecosystem | Next.js, React, Remix | | Vue ecosystem | Vue, Nuxt | | Other frontend | Svelte, SvelteKit, Angular | | Node.js backend | Express, Fastify, NestJS, Hono | | Python | Django, Flask, FastAPI | | Ruby | Rails, Sinatra | | Rust | Actix Web, Axum, Rocket | | Go | Gin, Echo, Fiber, Gorilla | | PHP | Laravel, Symfony |

Testing

| Tool | Ecosystem | |------|-----------| | Vitest | Node.js | | Jest | Node.js | | Mocha | Node.js | | Playwright | Browser | | Cypress | Browser | | pytest | Python | | Go test | Go | | Rust built-in | Rust | | RSpec | Ruby | | PHPUnit | PHP |

Linters & Formatters

| Tool | Type | Ecosystem | |------|------|-----------| | ESLint | Linter | JS/TS | | Prettier | Formatter | JS/TS | | Biome | Both | JS/TS | | Ruff | Both | Python | | Black | Formatter | Python | | Pylint / Flake8 | Linter | Python | | Clippy | Linter | Rust | | rustfmt | Formatter | Rust | | golangci-lint | Linter | Go | | RuboCop | Both | Ruby |

Databases & ORMs

| Database | ORM | |----------|-----| | PostgreSQL | Prisma, Drizzle, TypeORM, Sequelize | | MySQL | TypeORM, Sequelize | | SQLite | Prisma, Drizzle | | MongoDB | Mongoose | | Redis | — | | — | SQLAlchemy (Python), Diesel (Rust), GORM (Go), ActiveRecord (Ruby) |

Infrastructure

| Category | Detected | |----------|----------| | Deployment | Docker, Vercel, Netlify, Fly.io, Railway, Render, AWS CDK, Kubernetes, Cloudflare Workers, Heroku, Google Cloud | | CI/CD | GitHub Actions, GitLab CI, CircleCI, Jenkins, Travis CI, Bitbucket Pipelines | | Monorepo | Turborepo, Nx, Lerna, pnpm workspaces, npm/yarn workspaces, Cargo workspaces | | API Style | REST, GraphQL, gRPC, tRPC, WebSocket |

CLI Reference

Usage: dotai [options]

Options:
  -f, --format <formats>  Comma-separated: claude,cursor,windsurf,copilot,codex (default: all)
  -i, --interactive        Guided setup with prompts
  -o, --output <dir>       Target directory (default: current directory)
  -d, --dry-run            Preview output without writing files
  --force                  Overwrite existing config files
  -V, --version            Show version number
  -h, --help               Show help

Examples

# Generate all configs in current directory
npx dotai

# Only Claude and Cursor
npx dotai -f claude,cursor

# Preview what would be generated
npx dotai --dry-run

# Point at another project
npx dotai -o ~/projects/my-api

# Overwrite existing configs
npx dotai --force

# Interactive mode — confirm detections, select outputs
npx dotai -i

Why dotai?

The AI coding assistant landscape is fragmented. Claude Code reads CLAUDE.md. Cursor reads .cursorrules. Copilot reads .github/copilot-instructions.md. Each tool has its own format, its own conventions, its own location.

Most developers either:

1. Don't configure any of them (leaving AI suggestions generic and unhelpful)
2. Configure one and ignore the rest
3. Manually maintain 5 different files that say roughly the same thing

dotai solves this by analyzing your actual codebase — not asking you to fill out a form — and generating all five files with accurate, up-to-date information about your stack.

It's:

• Zero config — no setup, no config file for the config file generator
• Zero cost — no API calls, no tokens, no subscriptions
• Zero network — works completely offline, on an airplane, in a submarine
• Fast — pure static analysis, typically under 2 seconds
• Accurate — reads your actual lock files, config files, and dependency declarations

How It Works

1. Scan — Collects the file tree using fast-glob (respects .gitignore, skips node_modules/vendor/dist)
2. Detect — Runs 10 independent detectors in parallel: language, package manager, framework, testing, linter/formatter, database/ORM, deployment, CI/CD, monorepo, API style
3. Generate — Passes the detection results through format-specific generators that produce properly structured output for each tool

Everything is static analysis. It reads filenames, config files, and dependency manifests. Nothing is executed, uploaded, or sent anywhere.

Contributing

Contributions welcome! Especially:

• New framework/tool detections
• Improved output templates
• Support for additional AI coding tools

git clone https://github.com/avinashchaubey/dotai.git
cd dotai
npm install
npm run build
npm test

# Test against any project
npm run dev -- --dry-run -o ~/some-project

Use conventional commits: feat:, fix:, refactor:, docs:, test:, chore:.

License

MIT — Avinash Chaubey