Tailscale Compute MCP Server

What this does

Tailscale Compute exposes a set of tools to Claude over the Model Context Protocol. After you add it to claude_desktop_config.json (snippet above) and restart Claude Desktop, those tools become callable inside any conversation.

Requirements

This server declares the following environment variables. Set only the ones the maker requires for your setup; this listing does not flag any of them as required:

TAILSCALE_COMPUTE_HOST
TAILSCALE_COMPUTE_LOCAL_ROOT
TAILSCALE_COMPUTE_REMOTE_ROOT
TAILSCALE_COMPUTE_REMOTE_SHELL
TAILSCALE_COMPUTE_CONNECT_TIMEOUT_SECONDS
TAILSCALE_COMPUTE_AUDIT_LOG

Set these via the env object in your MCP config (see claude_desktop_config.json snippet above).

How do I install Tailscale Compute?

Run the Claude Code command in the install section to add Tailscale Compute to Claude Code, or copy the ready-made config for your client from the install section on this page. Restart your client afterwards and the tools appear in your next conversation.

Which clients work with Tailscale Compute?

This page carries setup and config for Claude Code, Claude Desktop, Cursor and VS Code. Any other MCP client that accepts a standard server config can connect to Tailscale Compute as well.

Does Tailscale Compute need an API key?

This listing does not flag an API key as required. The server can use these environment variables: TAILSCALE_COMPUTE_HOST, TAILSCALE_COMPUTE_LOCAL_ROOT, TAILSCALE_COMPUTE_REMOTE_ROOT, TAILSCALE_COMPUTE_REMOTE_SHELL, TAILSCALE_COMPUTE_CONNECT_TIMEOUT_SECONDS, TAILSCALE_COMPUTE_AUDIT_LOG; the maker's docs say which of them your setup needs. They go in the env block of your MCP config.