@meeseeks-sdk/mcp
An MCP server that runs a quality gate on generated code, executing it in a sandbox and scoring it before returning the code to the user.
README
@meeseeks-sdk/mcp
MCP server for Claude Code — makes the AI verify its own code before giving it to you.
What it does
Instead of Claude Code giving you the first code it generates, this MCP runs a quality gate:
Claude Code calls quality_gate tool
↓
LLM generates code
↓
Sandbox executes it (real Node.js subprocess)
↓
Score 0-10 based on correctness + efficiency
↓
Score ≥ 8? → return verified code
Score < 8? → feed failures back to LLM → retry
You get code that already survived real execution — not just what looked good at first glance.
Install
npm install -g @meeseeks-sdk/mcp
Setup with Claude Code
claude mcp add meeseeks --scope user npx @meeseeks-sdk/mcp
That's it. The quality_gate tool is now available in all your Claude Code sessions.
Tell Claude when to use it
Add to your project's CLAUDE.md or ~/.claude/CLAUDE.md:
## Code Generation
When writing isolated utility functions or algorithms (any language),
use the meeseeks__quality_gate tool instead of writing directly.
Use mode: "fast" if the pattern was used before, "balanced" for new tasks.
For Dart/Flutter/Python code: always pass harness: "free".
For JS/Node.js utilities: omit harness (auto-detected).
Tool: quality_gate
| Parameter | Type | Default | Description |
|---|---|---|---|
task |
string |
required | What to build — be specific about function signature and export format |
mode |
fast|balanced|quality |
fast |
fast=1 iter, balanced=3, quality=5 |
harness |
string |
auto-detected | js-api, js-lrucache, js-ratelimiter, js-circuitbreaker, js-promisepool, js-maze, js-sudoku, js-wordle, free |
projectContext |
string |
— | Stack description for the judge: "TypeScript + Express + PostgreSQL" |
minScore |
number |
8 |
Stop when score ≥ this value |
Requirements
- Node.js 18+
- AWS credentials configured (
~/.aws/credentials) for Bedrock adapter - Or set
AWS_REGIONenvironment variable
Knowledge inheritance
The MCP stores winning strategies in ~/.meeseeks/memory.db (SQLite). Next time you ask for a similar function, the proven approach is injected into the prompt — better first attempts over time.
Powered by
@meeseeks-sdk/core — the quality gate engine extracted from Meeseeks Hive.
License
MIT
推荐服务器
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。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。