Repository Inspector MCP Server
Enables AI clients to inspect Git repository changes, run allowlisted validation commands, and generate Markdown review reports via MCP, with validation disabled by default for security.
README
Repository Inspector
This is a small TypeScript developer tool that inspects changes in a Git repository, runs optional validation commands, and produces a Markdown report. It can be used from a command line or exposed to AI clients through MCP.
Your task
Investigate the repository and improve it as you judge best. The starter works for a narrow happy path, but production use may expose correctness, safety, reliability, contract, output, documentation, or testing weaknesses.
You are not expected to finish everything. We care about how you investigate, prioritize, implement, verify, and explain a meaningful scope.
Product decision
This tool may be used directly by developers and by AI coding agents. Decide whether its production interface should be CLI-first, MCP-first, or hybrid. Implement improvements consistent with your decision.
There is no preferred label. Explain:
- The primary user and execution environment you assumed.
- The trust boundary and allowed capabilities.
- Reliability, discoverability, latency/context, and output-size tradeoffs.
- How the interfaces you continue to advertise stay behaviorally consistent.
- What evidence would change your decision.
Time and rules
- Maximum 90 focused minutes within 48 hours of receiving the invitation.
- Use AI coding tools freely. Verify their work and document at least one suggestion you corrected or rejected.
- Work in your own repository created from this template.
- Commit as you work and complete
SUBMISSION.mdin your final commit. - Completion is not required. Accurate scope and verification matter more than a large diff.
Setup
npm install
npm run typecheck
npm test
CLI
npm run inspector -- review --repo ./path/to/repo --format markdown
npm run inspector -- review --repo ./path/to/repo --validate "npm test"
The report is written to review-report.md.
MCP
Start the stdio server with:
npm run mcp-server
It exposes a review_repository tool taking repo_path (required), base_ref
(optional), and validation_commands (optional array of shell commands).
MCP arguments can be steered by content inside the repository under review (prompt injection), so running validation commands over MCP is disabled by default. To enable it, the server operator must set both of these environment variables — setting only one leaves validation disabled:
INSPECTOR_ALLOW_VALIDATION=1INSPECTOR_VALIDATION_ALLOWLIST="npm test,npm run typecheck"— a comma-separated list of the exact commands permitted to run; anything else is skipped.
When validation is disabled or a command isn't on the allowlist, the tool
still returns a full read-only report; the skipped command is reported with
status: skipped and a reason instead of being silently dropped.
Project layout
src/core.ts shared review orchestration
src/cli.ts command-line adapter
src/mcp-server.ts MCP adapter
src/git.ts Git inspection
src/validation.ts validation execution
src/report.ts Markdown report generation
test/ public starter tests
When finished, submit via Security → Report a vulnerability on this
repo — see SECURITY.md for exactly what to include. Do not reply by email;
that submission channel is not monitored.
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。