Proof Publisher MCP
This MCP server publishes commit-bound proof summaries to a specified HTTPS webhook, featuring fail-closed writes and a clear authority boundary for safe external actions.
README
Proof Publisher MCP
A deliberately small TypeScript MCP server that publishes commit-bound proof summaries to an HTTPS webhook. The project exists to make one consequential external write easy to inspect, test, and place behind an authority boundary.
Consequential boundary
The publish_proof MCP tool calls ProofPublishingService.publish(). The
service validates and normalizes the exact target, repository, full commit SHA,
and summary, then calls ProofPublisher.send(action). The live implementation's
only external write is the fetch() POST inside
HttpWebhookProofPublisher.send().
MCP publish_proof
-> normalize exact action + SHA-256 fingerprint
-> ProofPublisher.send(action) <-- authority seam
-> fetch(allowed HTTPS origin) <-- consequential provider call
Writes fail closed unless ENABLE_WEBHOOK_WRITES=true, and even then the
target must match WEBHOOK_ALLOWED_ORIGIN. Tests inject a recording provider;
they do not make network calls.
Run the checks
npm install
npm run ci
Run the MCP server
Read-only/fail-closed mode is the default:
npm run build
npm start
To permit a real webhook write, an operator must explicitly configure both the write switch and the only allowed origin:
ENABLE_WEBHOOK_WRITES=true \
WEBHOOK_ALLOWED_ORIGIN=https://hooks.example.com \
npm start
No credentials are stored in this repository. The project is intentionally small so future approval logic can bind to the normalized action immediately before the provider runs, without moving validation or external I/O into the authority layer.
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。