CleanAI

Clean your code with AI

🌐 Website  •  🚀 Get Started Free

🪄 What is CleanAI?

CleanAI finds and removes dead code from your projects — unused functions, stale imports, deprecated exports — so you can ship cleaner, faster code.

Works with: VS Code, Cursor, and any VS Code-based editor.

💀 The Problem

Every codebase accumulates dead weight:

• Functions nobody calls anymore
• Imports that do nothing
• Code that "might be needed later" (it won't)
• Legacy utils everyone's afraid to delete

This slows your builds, confuses your team, and makes refactoring terrifying.

✅ The Solution

CleanAI scans your entire workspace and tells you exactly what's safe to delete.

One click → Clean code.

✨ Features

⚡ Agent (Local) setup

CleanAI’s Agent (Local) runs on your machine via Ollama and can propose/apply edits with a red/green diff preview. When available, the Agent also includes generated CleanAI context (for example .cursor/rules/cleanai-context/RULE.md and .cleanai/docs/index.md) as prompt context to improve answers.

1. In VS Code/Cursor:
   • Open the CleanAI sidebar and click Agent (Local), or
   • Run Clean AI: Open Agent (Local) (opens the panel’s Agent tab)
2. CleanAI will automatically set up the local runtime:
   • If Ollama isn’t installed, the Agent tab will show a Get Ollama link
   • If the model isn’t downloaded yet, CleanAI will download it for you (default: qwen2.5-coder:7b)
   • If Ollama is installed but not running, CleanAI will try to start the server automatically
   • If anything fails, use Retry and expand Diagnostics in the Agent tab to copy debug info

Optional (terminal):

ollama pull qwen2.5-coder:7b

Settings (VS Code → Settings → Clean AI):

• cleanai.agenticChat.ollama.baseUrl (default: http://127.0.0.1:11434)
• cleanai.agenticChat.ollama.model (default: qwen2.5-coder:7b)
• cleanai.agenticChat.ollama.autoStartServer.enabled (default: true)
• cleanai.agenticChat.ollama.autoInstall.enabled (default: true, best-effort on macOS via Homebrew)

📥 Installation

VS Code / Cursor

1. Open the Extensions panel (Cmd+Shift+X)
2. Search for CleanAI
3. Click Install

💡 Tip: If you don't see it right away, scroll down in the results — we're there!

Command Line

code --install-extension cleanai.cleanai

📊 How It Works

💡 Tip: After you finish GitHub sign-in in your browser, return to VS Code / Cursor — you’ll see a “Signed in as …” notification.

That's it. No configuration, no complex setup.

💰 Pricing

→ Start Free  •  → Go Pro  •  → Enterprise

⌨️ Keyboard Shortcuts

🎯 Perfect For

• Growing startups — Keep velocity high as code scales
• Legacy codebases — Finally clean up years of tech debt
• Pre-release audits — Ship smaller bundles
• Engineering teams — Maintain code quality standards

🔒 Privacy & Security

• All analysis happens locally (tool-backed)
• Your code does not get uploaded as part of analysis
• AI chat requests are proxied via cleanai.pro (do not paste secrets)
• We only track scan counts, not code content
• SOC 2 compliance (Enterprise)

⚙️ Settings

• cleanai.explorerDecorations.enabled (default: true): Toggle purple U badges in the Explorer for files with unused-code findings.
• cleanai.cursorExplorer.gitUntrackedColorFix.enabled (default: true): Cursor-only: restore Git "untracked" Explorer color to green to avoid conflicting with CleanAI purple.
• cleanai.debug.webview.enabled (default: false): Debug webview issues by sending webview errors (and optional debug events) to the Clean AI Output channel (and Telegram if enabled).
• VS Code / Cursor global settings that can hide all file badges/colors:
  • explorer.decorations.badges
  • explorer.decorations.colors

📱 Telegram Logging

Send CleanAI logs to a Telegram channel for monitoring and debugging.

Setup

1. Create a Telegram Bot:

   • Open Telegram and search for @BotFather
   • Send /newbot and follow the prompts
   • Save the bot token (format: 123456789:ABCdef...)

2. Get your Chat ID:

   • Create a channel or group
   • Add your bot as an administrator
   • For public channels: use @your_channel_name
   • For private channels: forward a message to @userinfobot to get the ID

3. Configure VS Code Settings:

{
  "cleanai.telegram.enabled": true,
  "cleanai.telegram.botToken": "YOUR_BOT_TOKEN",
  "cleanai.telegram.chatId": "@your_channel_or_chat_id",
  "cleanai.telegram.minLogLevel": "Info"
}

Settings

Log Levels

• Debug: All logs (verbose, includes internal details)
• Info: Normal operations and progress updates
• Warn: Potential issues that don't stop execution
• Error: Only errors and exceptions

🧩 Swift / iOS notes (Periphery)

Periphery is the source of truth for Swift dead-code detection, but like any static analyzer it can report false positives for runtime / Interface Builder entrypoints.

False Positives (What Periphery Docs Say)

Periphery is excellent, but it is still a static analyzer. The official docs call out that false positives can happen depending on project structure (e.g. Objective‑C runtime usage, Codable synthesis, XCTest inheritance, Interface Builder limits, and multi-target scans).

CleanAI’s default behavior is to show raw Periphery output.

Optional Conservative Filtering

If you want fewer “gotchas” in the UI, you can enable:

• cleanai.swift.periphery.falsePositiveFiltering.enabled

This only filters high-confidence runtime entrypoint files (e.g. AppDelegate.swift, SceneDelegate.swift, main.swift). It does not apply message-based heuristics.

CleanAI also passes --retain-objc-accessible to Periphery to reduce false positives in mixed Swift/Objective‑C projects.

Optional AI Recommendations (Container + Scheme)

If your workspace has multiple .xcworkspace / .xcodeproj files (or many schemes), you can enable:

• cleanai.workspaceConfiguration.aiAssistance.enabled

This uses AI to recommend the best container/scheme, but you still confirm via the picker (it never auto-selects silently).

Manual Overrides

If a file is still incorrectly flagged:

• Ignore in CleanAI: Click Ignore in findings (writes to .cleanaiignore).
• Configure Periphery directly:
  • Use inline comment: // periphery:ignore:all
  • Retain files via: --retain-file <glob>
  • Use .periphery.yml for project-wide configuration

Multi-Target Projects

If you have multiple targets (app + framework, app + tests), ensure:

1. Select the correct scheme: Choose a scheme that builds all relevant targets
2. Use a workspace: If targets are in different projects, use .xcworkspace
3. Check target membership: Files may be used by targets not in the selected scheme

Reset Configuration

If you selected the wrong Xcode scheme/workspace/project:

• Run: CleanAI: Reset Swift Periphery Configuration
• Re-run Analyze to pick again

🧩 TypeScript notes (ts-prune)

In monorepos, TypeScript often has multiple tsconfig.json files (one per package/app). CleanAI will:

• Use the single tsconfig.json automatically when unambiguous
• Prompt you to pick tsconfig(s) when multiple candidates exist

Your selection is cached for the workspace so subsequent runs are stable.

🖥️ CLI Usage (Running Separately)

CleanAI can be run from the command line for CI/CD integration or batch analysis:

# Basic usage
node dist/cli/RunCleanAiAnalysisCli.js --workspace /path/to/project

# Swift project with specific Xcode container and scheme
node dist/cli/RunCleanAiAnalysisCli.js \\
  --workspace /path/to/project \\
  --swift-container MyApp.xcodeproj \\
  --swift-scheme MyApp

# Output as JSON for CI integration
node dist/cli/RunCleanAiAnalysisCli.js --workspace /path/to/project --json

# Verbose output for debugging
node dist/cli/RunCleanAiAnalysisCli.js --workspace /path/to/project --verbose

CLI Options

Requirements

• For Swift: Periphery must be installed (brew install peripheryapp/periphery/periphery)
• For Python: Vulture must be installed (pip install vulture)
• For TypeScript/JavaScript: No additional installation (uses npx)

🤝 Support

• 📧 team@beloyal.app
• 🌐 cleanai.pro
• 📖 Documentation

Stop maintaining dead code. Start shipping faster.

Get CleanAI →