boilerforge
Architecture governance MCP server for AI-built codebases, enabling health checks, template management, and project scaffolding with migration support.
README
boilerforge
Architecture governance wedge for AI-built codebases — doctor-first health checks and guardrails for production teams.
Boilerforge is now doctor-first. Instead of starting from a boilerplate registry, you start by checking architecture quality:
boilerforge doctor .
It scores your project, detects stack metadata, validates governance guardrails, and gives actionable recommendations.
Quickstart
npx -y @cometforge/boilerforge@latest doctor .
CI gate:
boilerforge doctor . --ci --min-score 70
Badge snippet:
boilerforge doctor . --badge
Example output:

Commands
boilerforge doctor [path]
Architecture health report with weighted score (/100).
Flags:
--jsonoutput machine-readable JSON--cifail with non-zero exit if below threshold--min-score <number>threshold for--ci(default 70)--badgeprint markdown badge snippet + score
boilerforge analyze [path]
Detect stack metadata:
- language/runtime
- framework
- package manager
- baseline project metadata
boilerforge validate [path]
Evaluate architecture guardrails and return:
- pass/fail
- score
- categorized issues (
critical,warning,info) - recommendations
boilerforge extract [path] [--name <template-name>]
Extract current project into local template registry:
- writes
.boilerforge-template.yamlin source project - creates template under
~/.boilerforge/registry/<template-name>/ - copies sane files only (excludes
.git,node_modules,dist, build artifacts, env secrets)
boilerforge create <template-name> [target-path]
Scaffold a project from local registry template and write lockfile (boilerforge.lock.json).
boilerforge upgrade [path]
Upgrade project based on lockfile + local template version:
- compares locked template version vs latest local registry
- applies migration hooks in order when present (
writeFile,appendFile,deleteFile,copyFromTemplate,patchJSON) - writes updated lockfile
Doctor Rule Engine (100 points)
- CI workflow exists — 12
- Env schema validation present — 10
- Typecheck script exists — 10
- Test script + test files exist — 12
- Lint configured — 10
- Security headers configured (web stacks) — 10
- Centralized logging present — 8
- Observability hook present (Sentry/OpenTelemetry/etc) — 8
- Dependency hygiene — 10
- Docs baseline (README + setup) — 10
Outputs include:
- score
/100 - detected stack summary
- categorized issues
- actionable recommendations
Local Registry Model
Registry root:
~/.boilerforge/registry
Template layout:
<template>/
template.yaml
files/
migrations/ (optional)
template.yaml accepts YAML (not just JSON), and is strictly validated for required fields (name, version, createdAt) with clear parse/validation errors.
Lockfile (boilerforge.lock.json) stores template name/version/source for deterministic upgrades.
Lifecycle Compatibility
Existing lifecycle workflows remain supported:
boilerforge init <project-name> [--workflow=claude]
boilerforge update [--dry-run]
boilerforge protect add <category> <name>
boilerforge protect remove <category> <name>
boilerforge status
MCP Server Tools (unchanged)
list_boilerplatesget_boilerplatesearch_boilerplatesscaffold_projectcheck_project_updates
Running the package without CLI args in non-interactive MCP context still starts the MCP server.
CI/CD
GitHub Actions workflows included:
ci.yml: install, lint, typecheck, test, build,doctor --ci --min-score 70, CLI smoke checks, and npm package integrity (npm pack) verification.publish.yml: release-driven publish pipeline (lint + typecheck + test + build + doctor gate + npm publish with provenance).
Minimal CI snippet for your own repos:
- name: Architecture gate
run: boilerforge doctor . --ci --min-score 70
doctor --ci exits non-zero when below threshold, so it can be used as a hard merge/release gate.
Development
git clone https://github.com/MRKT365-India/boilerforge.git
cd boilerforge
npm install
npm run build
npm test
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。