Tailscale MCP 802bbc81

MCPAI代理

提供方 sandraschi

AI代理 stdio · 本地进程 会话登录 社区 / 第三方 Claude DesktopClaude
查看接入方式

Tailscale MCP是一个FastMCP服务器,通过一组组合工具将Tailscale管理API暴露给AI助手和自动化工具,用于设备、DNS、服务和监控等管理任务,并提供可选的代理工作流和Web应用界面。

一句话选型结论

按资料字段整理

Tailscale MCP 802bbc81 属于AI代理类 MCP 服务器,stdio · 本地进程,会话登录,可直接接入 Claude Desktop、Claude,有可直接复制的远程端点配置。接入前先确认权限范围;涉及写入、删除或付款的操作,建议先在测试环境验证。

分类AI代理 传输方式stdio · 本地进程 鉴权会话登录 部署方式支持远程部署
GitHub 星标 3 仓库热度,无公开仓库时显示未标注
工具数 6 暴露给模型的工具数量
资源 / 提示词 0 / 0 资源与提示词模板数量
维护方 社区 / 第三方 核验等级 仅资料核对

工具与能力

AI代理
数量来自来源目录的标注,具体工具清单以官方文档为准
工具数量6暴露给模型可调用的工具个数
资源数量0可读取的资源条目
提示词数量0内置提示词模板数量

能力标签

支持客户端

适用场景

网络管理Python本地部署自动化工具TailscaleAPI集成AI助手

安装与配置

stdio · 本地进程 会话登录
客户端在本机拉起子进程,数据不出本机
uv run tailscale-mcp

这条命令只是入口,具体启动参数请按官方文档补齐,不要直接照搬到别的服务器。

  1. 1
    准备运行环境

    确认你的客户端支持远程端点,并拿到服务地址。

  2. 2
    安装服务

    复制上面的安装命令执行;命令里带占位符的位置,换成自己的密钥或路径。

  3. 3
    写入客户端配置

    在客户端的 MCP 配置里填好启动命令和参数,保存后完全退出客户端再打开,让新配置生效。

  4. 4
    验证是否可用

    连接成功后客户端会列出约 6 个工具,随便调一个只读工具确认能返回结果。

典型使用场景

3 类
按分类与标签整理的常见用法,具体以官方文档为准
给代理接外部能力

让代理在任务中途调用搜索、数据库等真实工具。

多步任务编排

把一串操作拆成步骤,由代理按顺序执行。

自动化重复流程

把固定流程交给代理跑,减少手工操作。

选择建议

选之前先看这几条
按条目字段给出的判断,不替你决定用不用

值得选的地方

  • 资料标注 6 个工具,职责相对集中,接入后不容易误操作。
  • 有现成的安装命令,不用自己找入口,照着跑就能起服务。
  • 已适配 Claude Desktop、Claude,换客户端不用重新研究配置。

先想清楚的地方

  • stdio 方式要求本地装好运行时(如 Node.js),换机器要重新装一遍。
  • 使用前要准备好会话登录对应的凭据,并按最小权限配置。
  • 版本更新与行为细节以官方仓库为准。

更适合

  • 需要在AI代理场景里让模型直接调用外部能力的人。
  • 已经在用 Claude Desktop、Claude 的用户,接入路径最短。
  • 想先快速试一下、能接受在本地或自有环境跑服务的人。

不太适合

  • 完全不想在本地装环境的人。
  • 不方便管理密钥或第三方授权的人。
  • 需要厂商级 SLA 和技术支持的生产环境。

同类对比

6 个同类
同分类里挑最接近的几条,字段以各自详情页为准
MCP 服务器工具数传输方式部署方式安装命令最后核验
Tailscale MCP 802bbc81当前 6 stdio · 本地进程 支持远程部署 有 2026-05-17
Filesystem 560c4363 13 未标注 本地部署 无 2026-05-17
Server Everything (everything) 11 stdio · 本地进程 本地部署 有 2026-05-17
Time 7365cb6d 2 SSE · 服务端推送 远程部署 有 2026-05-18
Everything Claude Code 未标注 stdio · 本地进程 本地部署 有 2026-05-17
Openmemory 557e1585 未标注 stdio · 本地进程 本地部署 有 2026-05-17

服务器档案

AI代理
字段来自项目仓库与公开索引,中文说明由本站标注
提供方sandraschi上游标注的维护方 sandraschi
分类AI代理本站按上游分类展示
连接方式stdio · 本地进程客户端在本机拉起子进程,数据不出本机
鉴权方式会话登录远程服务建议用最小权限凭据
维护方社区 / 第三方核验等级 仅资料核对
包名 / 运行时未标注 · Python本地安装时用到的包标识与运行时
部署类型支持远程部署上游给出的部署形态
最近核验2026-05-17上游最后核验时间

详细介绍

7.8 KB
来自项目仓库与上游目录的原文,未做删改

Tailscale MCP

Operate your tailnet from the AI tools you already use. Tailscale MCP is a FastMCP 3.1.0+ server that exposes the Tailscale Admin API to assistants and automation: devices, DNS, services, monitoring, Funnel, Taildrop-related flows, and more through a small set of portmanteau tools (many operations each, so your client does not drown in hundreds of tiny tool names). Optional SEP-1577 agentic workflows (run_agentic_tailnet_workflow) let a capable host run multi-step flows when sampling is configured. An optional Webapp (Vite/React) gives humans a glass-style dashboard: My tailnet topology, Partner tailnets insights, help, and connection status.

Repository: github.com/sandraschi/tailscale-mcp


Why this exists

If you manage a Tailscale tailnet, you already use the admin console and the official API. This project turns that surface into MCP tools so Cursor, Claude Desktop, Antigravity, or any MCP-capable client can list devices, inspect DNS, drive automation, and answer questions about your network with structured responses (success flags, errors, hints) instead of ad-hoc shell scripts.

You stay on the Admin API contract: the authoritative reference remains Tailscales API docs. The MCP server adds opinionated grouping, persistence for certain flows, optional LLM sampling for multi-step tasks, and an optional browser UI for quick visual checks.


What you get

MCP tools (portmanteau pattern)

Instead of dozens of one-off tools, operations are grouped into verb-led domains such as manage_tailnet_devices, manage_tailnet_network, monitor_tailnet, manage_taildrop, manage_funnel, and more (security, automation, reporting, integrations, etc.). Each tool takes an operation argument similar in spirit to other fleet servers that avoid tool explosion. MCP names are verb-first; the redundant tailscale_ prefix is not used because this server is Tailscale-only.

Coverage includes device and user management, MagicDNS and network policies, Tailscale Services (TailVIPs) where exposed, monitoring and metrics, Funnel and Taildrop-related flows (with CLI integration where applicable), plus help, status, and partner tailnet summaries for orgs that share tailnets with partners.

The full matrix of tools and operations lives in docs/TAILSCALE_MCP_PORTMANTEAU_TOOLS.md (see the doc index for any additional reference material).

Agentic workflows (optional)

For multi-step automation, run_agentic_tailnet_workflow uses FastMCP sampling (SEP-1577). You can point the server at a local OpenAI-compatible endpoint (for example Ollama) via TAILSCALE_SAMPLING_*, or let the host run the LLM with TAILSCALE_SAMPLING_USE_CLIENT_LLM=1. Details and safety notes are in docs/PRD.md.

Webapp (optional)

The Webapp under web_sota/ is a React/Vite front end with a dark, glass-style layout. Highlights:

  • My tailnet visualize your network (Mermaid topology from get_tailnet_status, plus a decorative Orbit view).
  • Partner tailnets summary of members vs shared users and devices grouped by login, aligned with the summarize_partner_tailnets tool.
  • Help environment and sampling variables, linked from the shell.

Ports for this repo follow the fleet adjacency convention (10820 frontend / 10821 backend by default); see docs/WEBAPP.md and web_sota/start.ps1.

LM Link (Tailscale + LM Studio)

The get_lm_link tool provides operational control over LM Link — the Tailscale + LM Studio partnership (Feb 2026) that creates an encrypted mesh for remote local LLMs:

Operation CLI equivalent Description
status lms link status --json Live peers, loaded models, link state
enable / disable lms link enable/disable Toggle LM Link on/off
set_device_name lms link set-device-name Rename this device on the link
set_preferred_device lms link set-preferred-device Route model requests to a preferred peer
info / readiness (static / Tailscale API) Setup docs and tailnet connectivity check

The LM Link webapp page (/lm-link) shows a live dashboard with peer lists,
loaded model badges, device rename, and preferred device selection. Requires the
lms CLI (bundled with LM Studio or via curl -fsSL https://lmstudio.ai/install.sh | bash).

Cross-repo: For model-level management, see local-llm-mcp
which also probes LM Link status for its provider health dashboard.

Persistence and observability

DiskStore (FastMCP 3.1.0) holds durable state for funnels, transfers, and preferences across restarts see docs/STORAGE_BACKENDS.md. Optional Prometheus / Grafana / Loki-style stacks are documented under docs/monitoring/ if you want full observability on your own infrastructure.

Skills for agents

When present, skills/TAILSCALE_EXPERT.md is exposed as resource://tailscale/skills so clients can load operator-focused guidance alongside tools.


Who it is for

  • Platform engineers wiring tailnet changes into IDEs and automation.
  • Operators who want a quick dashboard plus MCP for deeper queries.
  • Agents (with appropriate allowlists) that should plan multi-step API work instead of one-shot guesses.

Requirements

You need a Tailscale API key with access to your tailnet and the tailnet name the key is scoped to:

Variable Role
TAILSCALE_API_KEY Bearer token from the admin console
TAILSCALE_TAILNET Your tailnet identifier (as shown in the console / API)

Optional variables for sampling, HTTP transport, and logging are documented in docs/PRD.md and .env.example.

If you are new to Tailscale itself, read docs/WHAT_IS_TAILSCALE.md first it explains tailnet vs Admin API vs client in plain language.


Quick Start```powershell

git clone https://github.com/sandraschi/tailscale-mcp
cd tailscale-mcp
just


This opens an interactive dashboard showing all available commands. Run `just bootstrap` to install dependencies, then `just serve` or `just dev` to start.

### Manual Setup

If you don't have `just` installed:

## Install (quick)

Full options (Claude Desktop JSON, Docker, Webapp) are in **docs/INSTALL.md**. Minimal path with **uv**:```bash
git clone https://github.com/sandraschi/tailscale-mcp.git
cd tailscale-mcp
uv sync
uv run tailscale-mcp

Set TAILSCALE_API_KEY and TAILSCALE_TAILNET in your environment or .env before starting.


Documentation map

Topic Document
Install, env, clients, Webapp docs/INSTALL.md
Tailscale concepts docs/WHAT_IS_TAILSCALE.md
Funnel / Taildrop / Services / Peer Relays — what each does, when to use it docs/FEATURES.md
Product scope and sampling docs/PRD.md
Webapp routes and ports docs/WEBAPP.md
Architecture docs/ARCHITECTURE_AND_DESIGN.md
Portmanteau tools (deep dive) docs/TAILSCALE_MCP_PORTMANTEAU_TOOLS.md
Everything else docs/DOCUMENTATION_INDEX.md
Contributing (uv, Ruff, tests) CONTRIBUTING.md

🛡️ Industrial Quality Stack

This project adheres to SOTA 14.1 industrial standards for high-fidelity agentic orchestration:

  • Python (Core): Ruff for linting and formatting. Zero-tolerance for print statements in core handlers (T201).
  • Webapp (UI): Biome for sub-millisecond linting. Strict noConsoleLog enforcement.
  • Protocol Compliance: Hardened stdout/stderr isolation to ensure crash-resistant JSON-RPC communication.
  • Automation: Justfile recipes for all fleet operations (just lint, just fix, just dev).
  • Security: Automated audits via bandit and safety.

License

MIT see LICENSE.

常见问题

8 问
答案由本站依据条目字段整理,不引用上游原文
Tailscale MCP 802bbc81 怎么安装?

在客户端按下面的安装命令接入:uv run tailscale-mcp。装完在客户端配置里引用同一个包或端点。

Tailscale MCP 802bbc81 怎么在 Claude Desktop、Claude 里配置?

它的传输方式是stdio · 本地进程,在客户端的 MCP 配置里填好启动命令和参数,保存后完全退出客户端再打开,配置才会生效。

Tailscale MCP 802bbc81 收费吗?

本站不代收费用,也没有标注具体价格。是否收费取决于服务本身和背后的接口额度,正式使用前按官方说明确认。

Tailscale MCP 802bbc81 需要 API Key 吗?

资料标注需要会话登录认证,使用前准备好对应凭据,并按最小权限配置。

Tailscale MCP 802bbc81 支持哪些客户端?

资料标注支持 Claude Desktop、Claude;其他客户端只要能配置 MCP Server,也能用同样的命令或地址接入。

Tailscale MCP 802bbc81 能提供哪些能力?

资料显示约 6 个工具;能力标签包括 网络管理、API集成、Python、Claude。具体工具清单以官方文档为准。

Tailscale MCP 802bbc81 需要什么运行环境?

运行时为 Python;属于本地进程,需要在本机装好运行环境。

Tailscale MCP 802bbc81 是本地还是远程?

它是本地进程(stdio),由客户端在本机拉起。传输方式为stdio · 本地进程。

热门专题

按标签挑服务器
全部专题 →

服务器信息以作者仓库与官方文档为准;分类、传输方式与鉴权方式的中文说明为中文整理,最近核验 2026-05-17。