Sponsored by Deepsite.site

Ai Quality Gate

Created By
mustafacagria month ago
🚀 Kill the Junior AI Era. 🤖 Level up your AI code to Principal standards. No more sloppy lines or junior mistakes. Automated ESLint ✨ TypeScript 🔧 Prettier 🎨 SonarQube 🛡️ Security 🔒 Complexity 📊 in seconds. High-stakes quality, forced to ship only the best. ⚡🏆 Goodbye, bad code. Hello, Principal Engineer. 🚀✨
Content

AI Quality Gate

MCP Server for AI code quality automation.

npm version License: MIT


What It Does

AI writes code → calls quality_fix → Server fixes what it can → Reports remaining issues to AI.

Hybrid Approach:

  • Phase 1: ESLint + 627 rules + Prettier (~2-8s, always runs)
  • Phase 2: SonarQube Server (~30-60s, optional)

Important: ESLint vs Prettier

ToolSourceConfig
ESLintProject's config (if exists) or MCP's embedded.eslintrc.* / eslint.config.* / MCP embedded
PrettierProject's ownProject's prettier.config.mjs

ESLint rules are controlled by MCP for consistent quality gates. Prettier uses project's config so formatting matches project preferences.

Phase 1 Rule Coverage:

PluginRulesDescription
SonarJS201Security, bugs, code smells
Unicorn127Modern JS best practices
ESLint Core108JavaScript fundamentals
TypeScript-ESLint99TypeScript-specific rules
RegExp60Regex best practices
Import11Import/export rules
Promise10Async/await best practices
Node.js (n)9Node.js specific rules
Unused Imports2Auto-remove unused imports
Total627

Installation

Prerequisites

  • Node.js 18+ on your PATH (node -v).
  • Cursor (or another MCP-capable editor) with MCP enabled.

Project root is auto-detected when PROJECT_ROOT is omitted: the server walks up from the MCP process working directory until it finds package.json or tsconfig.json. Set PROJECT_ROOT in env only to analyze a different tree than the inferred root.


MCP configuration (Cursor)

Open Settings → Tools & MCP → Edit (user mcp.json). Add one server block; the examples below match .cursor/mcp.json.example (JSONC with comments — if your editor rejects comments, copy the JSON blocks below only).

Server name vs tool name: The key under mcpServers (e.g. "ai-quality-gate") is only the label for that connection in Cursor. The MCP tool your agent calls is always quality_fix — that name is fixed by this package and is separate from the server key and from ai-quality-gate.

Always runs the published package; good for teams and CI-like setups.

{
  "mcpServers": {
    "ai-quality-gate": {
      "command": "npx",
      "args": ["-y", "ai-quality-gate"]
    }
  }
}

B) Optional: global npm install

After npm i -g ai-quality-gate, the ai-quality-gate binary is on your PATH:

{
  "mcpServers": {
    "ai-quality-gate": {
      "command": "ai-quality-gate",
      "args": []
    }
  }
}

C) SonarQube (Phase 2)

Requires a running SonarQube instance, sonar-scanner available (see SonarQube Setup), and all three variables below. Phase 1 still runs first.

{
  "mcpServers": {
    "ai-quality-gate": {
      "command": "npx",
      "args": ["-y", "ai-quality-gate"],
      "env": {
        "SONAR_HOST_URL": "http://localhost:9000",
        "SONAR_TOKEN": "your_sonar_token",
        "SONAR_PROJECT_KEY": "your_project_key"
      }
    }
  }
}

D) Optional environment variables (any server)

Add an "env" object when you need overrides. Merge order for config is defaults → .quality-gate.yaml / .quality-gate.json → environment variables.

VariableWhen to set
QUALITY_GATE_CONFIGAbsolute path to a specific .quality-gate.yaml or .quality-gate.json (skips walking directories).
PROJECT_ROOTForce project root if auto-detection is wrong for your layout.
SONAR_HOST_URLSonarQube server URL (with Phase 2).
SONAR_TOKENSonarQube token (with Phase 2).
SONAR_PROJECT_KEYSonarQube project key (with Phase 2).
SONAR_SCANNER_PATHFull path to sonar-scanner if not on PATH.
PHASE1_TIMEOUTPhase 1 timeout (ms), default 30000.
PHASE2_TIMEOUTPhase 2 timeout (ms), default 300000.
ENABLE_I18N_RULEStrue / false — stricter JSX literal checks for i18n projects.

Local development (this repository)

To dogfood or contribute:

  1. yarn build — generates dist/server.js.
  2. Point MCP at the built file (absolute paths):
{
  "mcpServers": {
    "ai-quality-gate-dev": {
      "command": "node",
      "args": ["/ABSOLUTE/PATH/TO/ai-quality-gate/dist/server.js"]
    }
  }
}
  1. Reload MCP. Use env.PROJECT_ROOT only if the repo you analyze differs from the inferred root.

2. Add AI Rule

SettingsRules and CommandsAdd Rule:

After every code change, before telling the user "done",
AI must call the quality_fix MCP tool. This is mandatory.

3. Use It

AI writes code → calls quality_fix → Fixes errors → "Done ✅"

CLI: interactive config (--setup)

The interactive wizard creates or updates .quality-gate.yaml without hand-editing: it walks you through project root, optional SonarQube (host URL + project key; token is not saved to disk — use SONAR_TOKEN in your environment), which Phase 1 tools to enable (ESLint, curly-brace / arrow AST fixers, Prettier, JSON validator), timeouts, and i18n rules. The generated file includes a fixers: block you can adjust later.

After yarn build (or install from npm), run from the target project (or any path under it):

node dist/server.js --setup

PROJECT_ROOT is inferred when unset (see MCP configuration). Use the same entrypoint as the MCP server (node dist/server.js or npx ai-quality-gate); only the --setup flag switches to wizard mode. Answer prompts in the terminal; on success you get a ready-to-use config next to your project root.

Other CLI modes: --check (read-only Phase 1), --fix (default behavior when using CLI quality run), --phase1-only, --phase2-only — see docs/DEVELOPMENT.md.


Optional: SonarQube Server (Phase 2)

Configure Sonar env in MCP as in C) SonarQube (Phase 2) above, or copy from .cursor/mcp.json.example. You need sonar-scanner on your machine for analysis (see below).

SonarQube Setup

# Start SonarQube
docker run -d --name sonarqube -p 9000:9000 sonarqube:community

# First login: admin/admin → change password
# http://localhost:9000

Docker Compose

# docker-compose.yml
version: '3'
services:
  sonarqube:
    image: sonarqube:community
    ports:
      - '9000:9000'
    volumes:
      - sonarqube_data:/opt/sonarqube/data
      - sonarqube_logs:/opt/sonarqube/logs
      - sonarqube_extensions:/opt/sonarqube/extensions

volumes:
  sonarqube_data:
  sonarqube_logs:
  sonarqube_extensions:
docker-compose up -d

Creating SonarQube Token

  1. http://localhost:9000 → Login (admin)
  2. My AccountSecurityGenerate Tokens
  3. Select token type: Global Analysis Token
  4. Copy token → use as SONAR_TOKEN

Installing sonar-scanner

PlatformMethodCommand
Windowsnpm (global)npm install -g sonarqube-scanner
WindowsChocolateychoco install sonar-scanner
macOSnpm (global)npm install -g sonarqube-scanner
macOSHomebrewbrew install sonar-scanner
Linuxnpm (global)npm install -g sonarqube-scanner
DockerContainerdocker run sonarsource/sonar-scanner-cli

For custom path: SONAR_SCANNER_PATH env var


Configuration

Optional files (discovered by walking up from the inferred project root — same algorithm as package.json / tsconfig.json — or from PROJECT_ROOT when set): .quality-gate.yaml (preferred) or .quality-gate.json. Same fields as environment variables (camelCase); you may nest Sonar settings under sonar: { hostUrl, token, projectKey, scannerPath }.

Merge order: defaults → config file → environment variables (ENV wins on conflicts).

Set QUALITY_GATE_CONFIG to an explicit path to skip discovery.

Custom rules (customRules)

Optional line-based regex checks on lintable files (Phase 1). Each match is reported as an issue with rule set to custom:<id> (and included in quality_fix remaining). Example:

customRules:
  - id: no-console
    message: 'Console.log is not allowed'
    pattern: 'console\\.log\\('
    severity: error
  - id: no-debugger
    message: 'Debugger statement found'
    pattern: 'debugger'
    severity: warning

Patterns use JavaScript RegExp source (escape backslashes as in YAML strings). Invalid patterns are skipped at runtime with a log line.

JSON validator & i18n locale files

When fixers.jsonValidator is enabled and you pass JSON paths that match locale patterns (for example locales/en.json / locales/tr.json), the tool compares keys across those files.

  • Syntax errors, invalid UTF-8 BOM, etc. → reported as issues and fail Phase 1 / quality_fix until fixed.
  • Missing or extra keys between locale files → collected as i18nIssues in the validator result and printed as warnings on stderr during Phase 1. They do not set passed: false and do not block the gate.

Treat i18nIssues as advisory unless you add your own CI check on top.


Environment Variables

All variables are optional unless you use Phase 2, which requires SONAR_HOST_URL, SONAR_TOKEN, and SONAR_PROJECT_KEY together.

VariableDescriptionExample
QUALITY_GATE_CONFIGAbsolute path to a .quality-gate.yaml or .quality-gate.json file. Skips walking parent directories for config discovery./app/ci/quality-gate.yaml
PROJECT_ROOTOverride detected project root. Default: walk up from the process cwd until package.json or tsconfig.json is found./Users/me/my-repo
SONAR_HOST_URLSonarQube server base URL (Phase 2).http://localhost:9000
SONAR_TOKENSonarQube authentication token (Phase 2). Prefer env / secret store; avoid committing.sqa_xxx...
SONAR_PROJECT_KEYSonarQube project key (Phase 2).my-project
SONAR_SCANNER_PATHFull path to the sonar-scanner executable if it is not on PATH./opt/sonar-scanner/bin/sonar-scanner
PHASE1_TIMEOUTPhase 1 subprocess timeout in milliseconds.30000 (default)
PHASE2_TIMEOUTPhase 2 (Sonar) timeout in milliseconds.300000 (default)
ENABLE_I18N_RULESSet to true to enable ESLint rules that flag raw string literals in JSX (for i18n-heavy apps).false (default)

Auto-Fix

Phase 1 automatically fixes these issues:

ESLint Auto-Fix (~100+ rules)

// var → const/let
var x = 1const x = 1

// forEach → for...of (unicorn/no-array-for-each)
arr.forEach(x => f(x))for (const x of arr) f(x)

// Nested ternary → extracted (unicorn/no-nested-ternary)
a ? b : c ? d : e  →  const temp = c ? d : e; a ? b : temp

// Unused imports removed
import { unused } from 'x'    (removed)

// Type imports (consistent-type-imports)
import { Type } from 'x'import type { Type } from 'x'

// Optional chain (prefer-optional-chain)
a && a.b && a.b.c  →  a?.b?.c

// Regex optimization (regexp/*)
/[0-9]//\d/

AST Auto-Fix

// Remove unnecessary curly braces (single-line if)
if (x) { return true }if (x) return true

Prettier Formatting

After ESLint fixes, Prettier runs to ensure consistent formatting:

// ESLint removes braces but leaves awkward format:
if (x)
  return true

// Prettier fixes to single line:
if (x) return true

Note: Prettier uses project's config, not MCP's.

Everything else: Reported to AI, AI fixes it.


API

Tool: quality_fix

// Input
{
  files: string[] // File paths to check
}

// Output
{
  phase: "local" | "server" | "complete",
  success: boolean,
  message: string,
  fixed: {
    eslint: number,          // ESLint auto-fixes
    curlyBraces: number,   // AST: single-statement if braces
    singleLineArrow: number, // AST: arrow body style
    prettier: number,      // Prettier formatting
    json: number           // JSON validation passes counted
  },
  remaining: Issue[],
  timing: {
    phase1: string,
    phase2?: string,
    total: string
  }
}

Feature Flags

ENABLE_I18N_RULES

For projects with internationalization (i18n), enable literal string detection:

{
  "env": {
    "ENABLE_I18N_RULES": "true"
  }
}

When enabled:

// ⚠️ Warning
<h1>Hello World</h1>

// ✅ OK
<h1>{t('hello')}</h1>

Troubleshooting

MCP: quality_fix does not appear

  1. Node.js 18+ — run node -v and npx --version.
  2. Reload Cursor after editing mcp.json (or use the MCP refresh control).
  3. JSON — the file must be valid JSON (no trailing commas). Copy from the MCP configuration section if unsure.
  4. Global install — if you use "command": "ai-quality-gate", run npm i -g ai-quality-gate once so the binary exists.

Windows

"npx not found" error:

# Node.js must be in PATH
# Check in PowerShell:
where.exe npx

Permission denied:

# Run PowerShell as Administrator

macOS / Linux

"Permission denied" error:

# Fix npm global directory
mkdir ~/.npm-global
npm config set prefix '~/.npm-global'
echo 'export PATH=~/.npm-global/bin:$PATH' >> ~/.zshrc
source ~/.zshrc

"sonar-scanner not found" error:

# Install via Homebrew
brew install sonar-scanner

# Or via npm
npm install -g sonarqube-scanner

SonarQube

"Insufficient privileges" error:

  • SonarQube → AdministrationSecurityGlobal Permissions
  • Give Anyone group Browse and Execute Analysis permissions

"Project not found" error:

  • Create project manually for first analysis: ProjectsCreate ProjectManually

Clone & build (contributors)

git clone https://github.com/mustafacagri/ai-quality-gate.git
cd ai-quality-gate
yarn install
yarn build

Use Local development (this repository) for MCP pointing at dist/server.js.


Docs

  • SETUP.md — Local setup (if included in your tree)
  • AGREEMENTS.md, docs/ARCHITECTURE.md, etc. — optional; some files may be omitted in minimal clones. README + .cursor/mcp.json.example are enough to run the published package.

Principles

  • ✅ 627 ESLint rules (SonarJS, Unicorn, TypeScript-ESLint, etc.)
  • ✅ Prettier integration (uses project's config)
  • ✅ AST-based transforms (no regex)
  • ✅ Verify after each fix
  • ✅ Rollback on error
  • ✅ ESLint config discovery (uses project config if available, otherwise embedded)
  • ✅ Zero workaround
  • ✅ Principal level

License

MIT © Mustafa Çağrı Güven


v0.0.1 — Initial release! MCP quality_fix, Phase 1/2 pipeline, CLI, config files, custom rules (see CHANGELOG)

Server Config

{
  "mcpServers": {
    "ai-quality-gate": {
      "command": "npx",
      "args": [
        "-y",
        "ai-quality-gate"
      ]
    },
    "ai-quality-gate-custom-root": {
      "command": "npx",
      "args": [
        "-y",
        "ai-quality-gate"
      ],
      "env": {
        "PROJECT_ROOT": "/path/to/your/project"
      }
    },
    "ai-quality-gate-sonar": {
      "command": "npx",
      "args": [
        "-y",
        "ai-quality-gate"
      ],
      "env": {
        "PROJECT_ROOT": "/path/to/your/project",
        "SONAR_HOST_URL": "http://localhost:9000",
        "SONAR_TOKEN": "your_sonar_token",
        "SONAR_PROJECT_KEY": "your_project_key"
      }
    },
    "ai-quality-gate-global": {
      "command": "ai-quality-gate",
      "args": []
    }
  }
}
Recommend Servers
TraeBuild with Free GPT-4.1 & Claude 3.7. Fully MCP-Ready.
Playwright McpPlaywright MCP server
EdgeOne Pages MCPAn MCP service designed for deploying HTML content to EdgeOne Pages and obtaining an accessible public URL.
Howtocook Mcp基于Anduin2017 / HowToCook (程序员在家做饭指南)的mcp server,帮你推荐菜谱、规划膳食,解决“今天吃什么“的世纪难题; Based on Anduin2017/HowToCook (Programmer's Guide to Cooking at Home), MCP Server helps you recommend recipes, plan meals, and solve the century old problem of "what to eat today"
Baidu Map百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Serper MCP ServerA Serper MCP Server
WindsurfThe new purpose-built IDE to harness magic
AiimagemultistyleA Model Context Protocol (MCP) server for image generation and manipulation using fal.ai's Stable Diffusion model.
Zhipu Web SearchZhipu Web Search MCP Server is a search engine specifically designed for large models. It integrates four search engines, allowing users to flexibly compare and switch between them. Building upon the web crawling and ranking capabilities of traditional search engines, it enhances intent recognition capabilities, returning results more suitable for large model processing (such as webpage titles, URLs, summaries, site names, site icons, etc.). This helps AI applications achieve "dynamic knowledge acquisition" and "precise scenario adaptation" capabilities.
RedisA Model Context Protocol server that provides access to Redis databases. This server enables LLMs to interact with Redis key-value stores through a set of standardized tools.
Visual Studio Code - Open Source ("Code - OSS")Visual Studio Code
Jina AI MCP ToolsA Model Context Protocol (MCP) server that integrates with Jina AI Search Foundation APIs.
MCP AdvisorMCP Advisor & Installation - Use the right MCP server for your needs
CursorThe AI Code Editor
DeepChatYour AI Partner on Desktop
Y GuiA web-based graphical interface for AI chat interactions with support for multiple AI models and MCP (Model Context Protocol) servers.
ChatWiseThe second fastest AI chatbot™
Tavily Mcp
MiniMax MCPOfficial MiniMax Model Context Protocol (MCP) server that enables interaction with powerful Text to Speech, image generation and video generation APIs.
BlenderBlenderMCP connects Blender to Claude AI through the Model Context Protocol (MCP), allowing Claude to directly interact with and control Blender. This integration enables prompt assisted 3D modeling, scene creation, and manipulation.
Amap Maps高德地图官方 MCP Server