MCP服务器目录

Wedgwood Web Works 的 MCP 服务器目录与运维地图,汇总了档案、聊天桥接、终端自动化、CSS 分析、图表、图像处理、可观测性、邮件、内容发布、图数据库、酒店 PMS、设计自动化、DNS/域名、书签、抓包、3D 场景、预订和网络运维等 20 个 MCP 服务器的架构、常用命令、文档规范与安全说明,供开发者和运维人员在本地开发及 MCP 客户端接入时查阅。

  • by lesleslie
  • 本地部署
  • 1 星标
  • 更新于
在 GitHub 查看 lesleslie/www-mcp-servers

使用说明

Wedgwood Web Works MCP Servers

Catalog and operating map for the Wedgwood Web Works MCP server repositories.

Status: Internal Bodai ecosystem catalog

Quick Links

Quality & CI

Crackerjack is the standard quality-control and CI/CD gate across the MCP server repos. Individual repositories may still expose direct pytest, ruff, pyright, or mypy commands for targeted debugging, but repo-level validation should stay aligned with the Crackerjack workflow.


Overview

This repository is the catalog for the *-mcp servers in the Wedgwood Web Works and Bodai ecosystem. It documents the shared stack, lists the active server repos, and provides a common operator orientation for local development and MCP client wiring.

The catalog is not itself an MCP server package. Use it as the index when deciding which integration server owns a capability, which port it normally binds to, and which repo README should hold detailed setup and tool documentation.

Architecture

These servers use a layered architecture that keeps provider-specific code small and consistent:

  • FastMCP: tool and route registration for the MCP protocol surface
  • mcp-common: standard lifecycle CLI with start, stop, restart, status, and health
  • Oneiric: shared runtime patterns, structured logging, snapshots, and health conventions
  • Pydantic: typed request, response, and settings models
  • UV: dependency management and local command execution

Typical server shape:

<server>_mcp/
  cli.py        # mcp-common lifecycle CLI
  client.py     # provider API boundary, when needed
  config.py     # Pydantic settings and logging
  models.py     # typed domain models
  server.py     # FastMCP application factory
  tools/        # registered MCP tools
settings/
  <server>.yaml # committed defaults
tests/

MCP Servers

Server Role Default Port Description
archive-org-mcp Web archives 3054 Wayback Machine snapshots, Internet Archive catalog search, and saved-page retrieval
chat-bridge-mcp Cross-app bridge 3057 Bidirectional bridge between Claude Desktop and ChatGPT Desktop on the same machine via Chrome DevTools Protocol (forward, ask, and peer discovery tools)
cmux-mcp Terminal automation 3061 Programmatic control of a running cmux instance (macOS Ghostty terminal for AI agents) — orchestration and notifications over Unix-socket JSON-RPC plus browser automation via the cmux CLI
css-mcp CSS analysis 3050 CSS metrics, MDN docs, browser compatibility, and project analysis
excalidraw-mcp Diagramming 3032 Real-time collaborative Excalidraw scene and WebSocket workflows (canvas WebSocket on 3060)
graphics-mcp Image operations 3040 Raster image metadata, conversion, resizing, cropping, filters, and thumbnails
langsmith-mcp Observability 3048 LangSmith traces, prompts, datasets, experiments, and usage lookup
mailgun-mcp Email operations 3039 Mailgun email, domains, events, suppression lists, templates, and validation
medium-mcp Content publishing 3055 Medium article, author, and publication lookup via RapidAPI medium2
neo4j-mcp Graph database 3045 Cypher, node, relationship, path, and schema operations
opera-cloud-mcp Hospitality PMS 3037 Oracle OPERA Cloud reservations, guests, rooms, operations, and financial data
penpot-api-mcp Design automation 3051 Penpot REST API access — projects, files, and design object read/search/export without a browser
porkbun-dns-mcp DNS management 3042 Porkbun DNS record listing, creation, editing, and deletion
porkbun-domain-mcp Domain lifecycle 3043 Porkbun domain inventory, metadata, auth codes, renewal, and pricing
raindropio-mcp Bookmarks 3034 Raindrop.io bookmark, collection, tag, batch, and filtering operations
scapy-mcp Packet analysis 3056 Scapy packet crafting, dissection, and live capture
spline-mcp 3D scenes 3052 Spline.design scene orchestration and asset workflows
synxis-crs-mcp Reservations 3046 SynXis CRS property search, availability, rates, and booking
synxis-pms-mcp Property operations 3047 SynXis PMS guest, room, check-in, check-out, and folio workflows
unifi-mcp Network operations 3038 UniFi Network and Access controller management

Operating Model

The servers should feel interchangeable at the operational layer even when their provider APIs differ.

Expected common behavior:

  • uv sync --group dev installs local development dependencies
  • uv run <server-name> start starts the HTTP MCP server
  • uv run <server-name> health returns local health metadata
  • settings/<server>.yaml contains committed defaults
  • environment variables override committed defaults
  • provider credentials stay in environment variables, local .env files, or a secret manager
  • /health and /healthz are available on HTTP-enabled servers that follow the current template

Each plugin is distributed through this repo's Claude Code marketplace: register once with claude plugin marketplace add https://github.com/lesleslie/www-mcp-servers, then claude plugin install <server> per server to install as a Claude Code plugin alongside the mcp__<server>__* tools.

Common Commands

Run these from an individual server repo:

uv sync --group dev
uv run <server-name> start
uv run <server-name> status
uv run <server-name> health
uv run pytest
uv run ruff check <package_name> tests
uv run ruff format <package_name> tests

Some repos use pyright; others use mypy. Follow the individual README and pyproject.toml for the authoritative type-check command.

Documentation Standards

Each server README should include:

  • badges for Crackerjack, Oneiric, FastMCP, UV, and Python
  • a one-line purpose statement
  • version and status
  • quick links
  • Quality & CI notes
  • overview and capability bullets
  • quick start with mock mode when available
  • standard lifecycle CLI commands
  • MCP client configuration example
  • tool reference table
  • configuration table
  • project structure
  • development commands
  • security notes tailored to that provider

Keep catalog-level summaries here. Put detailed setup, provider-specific examples, and tool-level behavior in each server repo.

Security Notes

  • Never commit provider credentials, bearer tokens, customer data, or generated local .env files.
  • Treat mutating tools as privileged operations, especially DNS, domain renewal, email sending, database mutation, reservation creation, and hospitality check-in/check-out workflows.
  • Keep ports, hosts, provider URLs, tenant IDs, and account settings configurable.
  • Scrub real domains, email addresses, hotel data, guest data, graph data, and local paths before sharing examples outside local development.

Why WWW MCP Servers?

Most MCP servers in the wild are built directly on the MCP SDK, resulting in inconsistent patterns, missing production safeguards, and no standardized tooling. WWW MCP servers are built on the shared mcp-common foundation so each server gets the same operational model while keeping provider-specific logic isolated.

That gives the ecosystem:

  • Consistency: unified CLI, configuration patterns, and health behavior
  • Production posture: PID management, stale process recovery, graceful shutdown, and health persistence where supported
  • Security focus: credential masking, input validation, and narrow provider boundaries
  • Maintainability: improvements to shared lifecycle behavior can flow through the server fleet

Wedgwood Web Works