经 AI Skill Hub 精选评估,Scrypted MCP插件 获评「强烈推荐」。这款MCP工具在功能完整性、社区活跃度和易用性方面表现出色,AI 评分 8.0 分,适合有一定技术背景的用户使用。
Scrypted MCP插件 是一款遵循 MCP(Model Context Protocol)标准协议的 AI 工具扩展。通过 MCP 协议,它可以让 Claude、Cursor 等主流 AI 客户端直接访问和操作外部工具、数据源和服务,实现 AI 能力的无缝扩展。无论是文件操作、数据库查询还是 API 调用,都可以通过自然语言在 AI 对话中直接触发,极大提升生产效率。
Scrypted MCP插件 是一款遵循 MCP(Model Context Protocol)标准协议的 AI 工具扩展。通过 MCP 协议,它可以让 Claude、Cursor 等主流 AI 客户端直接访问和操作外部工具、数据源和服务,实现 AI 能力的无缝扩展。无论是文件操作、数据库查询还是 API 调用,都可以通过自然语言在 AI 对话中直接触发,极大提升生产效率。
# 方式一:通过 Claude Code CLI 一键安装
claude skill install https://github.com/sv-tools/scrypted-mcp-plugin
# 方式二:手动配置 claude_desktop_config.json
{
"mcpServers": {
"scrypted-mcp--": {
"command": "npx",
"args": ["-y", "scrypted-mcp-plugin"]
}
}
}
# 配置文件位置
# macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
# Windows: %APPDATA%/Claude/claude_desktop_config.json
# 安装后在 Claude 对话中直接使用 # 示例: 用户: 请帮我用 Scrypted MCP插件 执行以下任务... Claude: [自动调用 Scrypted MCP插件 MCP 工具处理请求] # 查看可用工具列表 # 在 Claude 中输入:"列出所有可用的 MCP 工具"
// claude_desktop_config.json 配置示例
{
"mcpServers": {
"scrypted_mcp__": {
"command": "npx",
"args": ["-y", "scrypted-mcp-plugin"],
"env": {
// "API_KEY": "your-api-key-here"
}
}
}
}
// 保存后重启 Claude Desktop 生效
A Scrypted plugin that exposes a Streamable HTTP Model Context Protocol endpoint, with OAuth backed by Scrypted user accounts. AI assistants (Claude Desktop, Claude Code, etc.) connect to your running Scrypted server and call MCP tools to inspect logs, manage plugins, query devices, and snapshot the database.
Search for scrypted-mcp in your Scrypted plugin browser, or install via the npm registry:
```bash
```
The plugin appears as MCP Server under your installed plugins. Click into it to see the endpoint URL in the console output on first boot.
The plugin's Settings tab in the Scrypted UI surfaces both configurable knobs and a live health snapshot. Open * Plugins → MCP Server → Settings* to see them.
| Setting | Default | Range | What it does |
|---|---|---|---|
dcr_max_clients | 100 | ≥ 1 | Cap on persisted DCR client registrations. When a /register would push the total above this, the least-recently-used registration is evicted. |
access_token_ttl_sec | 3600 (1 h) | 60 – 86400 | Lifetime of issued JWT access tokens. Shorter = tighter security; longer = less re-auth churn. |
refresh_token_ttl_sec | 2592000 (30 d) | 3600 – 31536000 | Lifetime of issued refresh tokens. Each refresh rotates the token (single-use), so this is the upper bound on how long a stolen RT remains valid. |
max_restore_mb | 500 | 1 – 10240 | Maximum decoded backup size that restore_backup will accept. Bump if your Scrypted database exceeds the default. |
All four values are read at request time, so changes take effect on the next inbound request — no plugin reload required.
Once installed and the plugin is running, the MCP endpoint is:
<your-scrypted-base-url>/endpoint/scrypted-mcp/public/mcp
Use whichever base you use for the Scrypted UI — https://host:10443 if you're on the secure port, or http://host:11080 if you're on the insecure one. The plugin auto-detects the protocol from the inbound port (or trusts x-forwarded-proto if you have a reverse proxy in front), so OAuth metadata documents come back with URLs your client can actually reach.
You don't need to point clients at any other URL — the protected-resource metadata at /endpoint/scrypted-mcp/public/.well-known/oauth-protected-resource advertises the authorization server, which advertises /authorize, /token, /register, and the JWKS document.
Two read-only URLs the plugin computes from the live Scrypted listening address — paste whichever one matches how you're reaching Scrypted.
- MCP endpoint URL (HTTPS) — for production clients and anything behind a reverse proxy. - MCP endpoint URL (HTTP) — for local-network clients and debugging where the self-signed HTTPS cert is awkward. Don't expose to the internet.
- list_plugins — All installed plugins with id / pluginId / name / type. - get_plugin_info — pid, pending RPC calls, object count for one plugin. - reload_plugin — Restart a plugin host (picks up code changes). - kill_plugin — Kill a plugin host without auto-reload. - install_plugin — Install / upgrade a plugin from npm. - update_plugins — Upgrade every installed plugin to its latest npm version. - npm_info — Query npmjs.org via the Scrypted server. Use "scrypted-kasa" for package metadata, or "-/v1/search?text=keywords:scrypted-plugin" to discover plugins. - rename_device_id — Rename a Scrypted device id (briefly kills the owning plugin). - set_mixins — Replace the list of mixins on a device. Read get_device first if you only want to add/remove one. - get_storage / set_storage — Read / replace a plugin device's persistent KV storage. - get_id_for_native_id — Reverse-lookup a Scrypted device id from a plugin's internal nativeId. - disconnect_clients — Disconnect all websocket clients of a plugin (forces reconnection). - clear_console — Clear a plugin device's console buffer.
If your MCP client logs HTTP 404 ... Cannot POST /register (or any other path at the host root), the plugin's OAuth metadata advertised URLs your client couldn't reach. Most often this means the protocol got mis-detected — check that your client's URL matches the port (https → 10443, http → 11080) and that nothing in front of Scrypted is rewriting Host without forwarding x-forwarded-proto.
After fixing the URL, re-run claude mcp remove scrypted and re-add it (or equivalent for your client) to clear any cached OAuth discovery state.
高质量的MCP插件,提供灵活的集成方案
AI Skill Hub 为第三方内容聚合平台,本页面信息基于公开数据整理,不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后,再部署至生产环境,并做好必要的安全评估。
✅ MIT 协议 — 最宽松的开源协议之一,可自由商用、修改、分发,仅需保留版权声明。
AI Skill Hub 点评:Scrypted MCP插件 的核心功能完整,质量优秀。对于Claude Desktop / Claude Code 用户来说,这是一个值得纳入个人工具库的选择。建议先在非生产环境试用,再逐步推广。
| 原始名称 | scrypted-mcp-plugin |
| 原始描述 | 开源MCP工具:Scrypted plugin that exposes a Streamable HTTP Model Context Protocol endpoint, 。⭐4 · TypeScript |
| Topics | mcpscryptedtypescript |
| GitHub | https://github.com/sv-tools/scrypted-mcp-plugin |
| License | MIT |
| 语言 | TypeScript |
收录时间:2026-06-14 · 更新时间:2026-06-14 · License:MIT · AI Skill Hub 不对第三方内容的准确性作法律背书。
选择 Agent 类型,复制安装指令后粘贴到对应客户端