🔍 RepoLens AI

The AI Context Intelligence Platform

Your AI agents are only as good as the context you give them.

Stop auto-generating garbage AGENTS.md. Start curating context that actually works.

The Problem

Auto-generated AGENTS.md files reduce AI agent performance by 2-3% (ETH Zurich, 2025). They waste tokens, add noise, and contain only information AI agents can already infer on their own.

Meanwhile, hand-curated context files with tribal knowledge improve performance by ~4%.

The problem isn't missing context files — it's bad context files.

What RepoLens AI Does

RepoLens AI is a context quality platform that helps you:

• ⭐ Setup AI context for 6 tools with ONE command
• Context — create a paste-ready project brief for ChatGPT, Claude, Gemini, Cursor, and Antigravity
• Prompt — turn vague ideas into project-aware AI coding prompts
• Check — sanity-check AI-generated changes before commit
• 📊 Score your context files quality (0-100, 5 dimensions)
• 🔄 Sync AGENTS.md → Cursor (.mdc) → Claude → Copilot → Windsurf → Codex
• 📚 Templates — 12 curated framework templates with real gotchas
• 🔧 Fix — auto-remove generic rules that hurt AI performance

For Antigravity & Vibe Coding

RepoLens does not replace Antigravity, Cursor, Claude Code, Codex, or Copilot. It prepares your repository so those agents stop guessing.

npx repolens-ai setup
npx repolens-ai context
npx repolens-ai prompt "add login with email verification"
npx repolens-ai check
npx repolens-ai vibe

setup creates high-signal project context. context gives any AI chat the short brief it needs. prompt turns an idea into a repo-aware instruction. check catches obvious AI-generated mistakes before they become project debt. vibe tells you whether the repo is ready for AI-assisted work and what to improve next.

Supported AI Tools

| Tool | File Generated | Format | |---|---|---| | Cursor IDE | .cursor/rules/project.mdc | MDC (modern, YAML frontmatter) | | Cursor IDE | .cursorrules | Legacy format | | Claude Code | CLAUDE.md | Markdown | | GitHub Copilot | .github/copilot-instructions.md | Markdown | | OpenAI Codex | CODEX.md | Markdown | | Windsurf | .windsurfrules | Rules format | | Antigravity / All | AGENTS.md | Universal (source of truth) |

Install

npm install -g repolens-ai
repolens --version

Requires: Node.js ≥ 18. Download at nodejs.org.

Quick Start

cd my-project

# ONE COMMAND — sets up everything
repolens setup

# Optional: see whether the repo is ready for AI-assisted work
repolens vibe

That's it. repolens setup will:

1. ✅ Analyze your project (framework, architecture, code)
2. ✅ Create AGENTS.md (with framework-specific template if available)
3. ✅ Sync to 6 AI tools (Cursor .mdc, Claude, Copilot, Windsurf, Codex)
4. ✅ Generate skill files for Cursor
5. ✅ Score your context quality

repolens vibe then gives you one community-readiness score across context quality, AI tool coverage, workflow clarity, and code health.

Want more control? Use individual commands:

repolens context       # Paste-ready context for any AI chat
repolens prompt "..."  # Project-aware prompt generator
repolens check         # Sanity-check AI-generated changes
repolens init          # Interactive AGENTS.md creation (10 questions)
repolens vibe          # Score vibe-coding readiness
repolens lint          # Score your context quality (0-100)
repolens fix           # Auto-fix generic rules
repolens sync          # Sync to all AI tools
repolens templates     # Browse 12 framework templates
repolens dashboard     # Visual web dashboard

Commands

Daily Vibe Coding Loop

Use these commands every day with Cursor, Antigravity, Claude, ChatGPT, Gemini, Codex, or any AI coding workflow.

repolens context
repolens prompt "add image upload to user profiles"
repolens check

repolens context

Creates a compact project brief that can be pasted into any AI chat. It includes the detected stack, architecture, key directories, critical files, rules, commands, and gotchas.

repolens context
repolens context --copy

repolens prompt "<request>"

Turns a vague feature idea into a project-aware prompt. RepoLens detects the task intent, finds the most relevant project files, suggests likely files to create or update, adds feature-specific guardrails, and gives the AI agent a clear verification plan.

repolens prompt "add customer avatar upload"
repolens prompt "add customer avatar upload" --copy

repolens check

Fast guardrail after AI edits. It scans changed files for risky changes, TypeScript any, missing use client in React hook files, and debug logging.

repolens check

🆕 Context Intelligence

repolens vibe

Score whether a repository is ready for vibe coding and community use. This is the fastest way to see if a repo has enough context, tool coverage, workflow clarity, and code-health signals for AI agents to work effectively.

repolens vibe

Checks:

• Context quality and drift
• Coverage for AGENTS.md, Claude/Codex, Cursor, Copilot, Windsurf, and skills
• Dev/build/lint/test scripts
• README, CI, license, and contribution readiness
• Security risks, circular dependencies, complexity, and documentation signals

repolens init

Interactive AGENTS.md creation through developer interview. Extracts tribal knowledge — gotchas, architecture decisions, specific commands — the stuff AI agents can't figure out on their own.

repolens init       # Interactive mode (recommended)
repolens init -y    # Auto-detect only, no questions

Asks smart questions like:

• "Any gotchas or pitfalls a developer should know?"
• "Key architecture decisions?"
• "Files that should NEVER be modified?"

repolens lint

Score and analyze your AI context files. Measures 5 dimensions:

repolens lint

Scores: | Metric | What It Measures | |---|---| | Specificity | Project-specific vs generic rules | | Coverage | Important areas covered | | Conciseness | Not too long (agents struggle with >200 lines) | | Freshness | References actual project files | | Tribal Knowledge | Non-inferable information (gotchas, decisions) |

Detects anti-patterns like:

• "Never commit secrets" → Generic, AI already knows this
• "Follow TypeScript conventions" → Generic, wastes tokens
• "Validate all input" → Generic, provides zero value

repolens sync

Keep all your AI context files in sync from a single source of truth.

repolens sync            # Sync AGENTS.md → all targets
repolens sync --dry-run  # Preview without writing
repolens sync --force    # Overwrite files with manual edits

Syncs to:

• CLAUDE.md (Claude Code)
• .cursorrules (Cursor IDE)
• .github/copilot-instructions.md (GitHub Copilot)

repolens doctor

Health check for your AI development setup.

repolens doctor

Checks: Context files · Git · .gitignore · Environment security · Scripts · Sync status

repolens dashboard

Launch a beautiful local web dashboard for visual context management.

repolens dashboard           # Open at http://localhost:3141
repolens dashboard -p 8080   # Custom port

Dashboard shows:

• Context quality scores with ring charts
• Risk analysis with severity breakdown
• Code intelligence (functions, classes, LOC)
• Context files sync status
• Largest files analysis

repolens templates

Browse and apply curated AGENTS.md templates for 12+ popular frameworks — each containing real tribal knowledge (gotchas, pitfalls, conventions) written by experienced developers.

repolens templates                    # Browse all templates
repolens templates --search react     # Search by keyword
repolens templates --apply nextjs     # Apply a template to your project

Available templates: Next.js, React+Vite, Vue/Nuxt, Express, NestJS, Django, FastAPI, Laravel, React Native, HTML/CSS/JS, T3 Stack, Supabase, and more.

repolens fix

Auto-fix detected issues — removes generic rules that hurt AI performance, cleans up boilerplate, and improves quality scores automatically.

repolens fix             # Fix and save (creates backups)
repolens fix --dry-run   # Preview fixes without writing

repolens skills

Generate .cursor/skills/ files — task-specific instructions that help AI agents perform complex tasks correctly (authentication, database, deployment, debugging, responsive UI).

repolens skills --list   # See available skills
repolens skills --all    # Generate all relevant skill files

Analysis & Review (Upgraded)

repolens analyze             # Full repo analysis (now reads code content)
repolens arch                # Architecture deep-dive
repolens explain auth        # Explain a module (now searches code, not just filenames)
repolens review              # Review recent changes
repolens risks               # Security & architecture risks
repolens onboard             # Generate onboarding guide

Why Not Just Write AGENTS.md Manually?

You should! But RepoLens AI helps you write better ones:

| Without RepoLens | With RepoLens | |---|---| | You forget to add gotchas | Interactive interview extracts tribal knowledge | | You don't know if your rules are generic | Lint scores specificity and detects anti-patterns | | You maintain 4 separate context files | Sync keeps everything in sync automatically | | You don't know if context is outdated | Freshness scoring detects stale references | | You can't visualize context quality | Dashboard shows scores and issues visually |

See a concrete before/after example of weak generic context versus context that actually helps AI agents.

Privacy First

• ✅ Never uploads code — everything runs locally
• ✅ Never reads .env values — detects existence only
• ✅ Masks all secrets in output
• ✅ Works fully offline — AI is optional
• ✅ No telemetry — zero tracking

Supported Frameworks

Auto-detects 15+ frameworks: Next.js, React, Vue, Angular, SvelteKit, NestJS, Express, Fastify, Hono, Laravel, Django, FastAPI, Flask, Rails, Go, Rust, Odoo, and more.

Project Structure

RepoLens keeps the source tree intentionally simple for contributors and AI coding agents:

src/
  cli.ts              CLI entrypoint and command registration
  commands/           One file per user-facing command
  core/               Analysis, scoring, sync, and workflow engines
  ai/                 Optional AI enhancement and privacy filtering
  reporters/          Terminal and markdown output
  templates/          Framework and tool context templates
  utils/              Shared logging, paths, masking, and error handling
tests/                Focused Vitest coverage for core behavior
docs/                 Command docs, privacy notes, roadmap
examples/             Before/after and sample context files

Build output lives in dist/ and is intentionally ignored in git. The npm package publishes the generated dist bundle after npm run build.

AI Enhancement (Optional)

Works perfectly without any API key. For AI-enhanced explanations:

export OPENAI_API_KEY=your-key-here
repolens analyze

Roadmap

• [x] Vibe readiness scoring command
• [x] Daily vibe coding commands: context, prompt, check
• [x] 18 CLI commands
• [x] Interactive knowledge extraction
• [x] Context quality scoring (5 dimensions)
• [x] Cross-tool sync (AGENTS.md → CLAUDE.md → .cursorrules → Copilot)
• [x] Local web dashboard
• [x] Code-content-aware analysis
• [x] Privacy-first design
• [ ] VSCode extension
• [ ] GitHub PR review bot
• [ ] Team collaboration dashboard
• [ ] Custom scoring rules
• [ ] Multi-repo analysis
• [ ] CI/CD integration (fail PR if context is stale)

Contributing

git clone https://github.com/Harry-Kien/repolens-ai.git
cd repolens-ai
npm install
npm run typecheck
npm test
npm run build
npm link

# Test locally
repolens context
repolens prompt "add login"
repolens check
repolens doctor

License

MIT © RepoLens AI Contributors

Built for the AI coding era.

Stop guessing. Start curating.

Get Started · Report Bug · Request Feature