CLI + MCP

The CLI can run as a local Model Context Protocol server over stdio, exposing your links as tools that AI agents can call. It uses your configured CLI credentials, so an agent acts as you against whichever server your active profile points at.

CLI MCP vs. hosted MCP server

This is different from the hosted MCP Server at mcp.lnkify.io, which speaks Streamable HTTP. The CLI's mcp serve runs locally over stdio — ideal for desktop agents like Claude Desktop that launch a local command. Both ultimately drive the same GraphQL API.

Run the server

lnkify mcp serve

Authentication comes from the same resolution chain as every other command — flags, env, the active profile, or the default. Make sure you've authenticated first (see Authentication).

Available tools

Tool	Description
create_link	Create a shortlink for a target URL
list_links	List your links with optional search/pagination
get_link	Get details for a single link by id
delete_link	Delete a link by id
resolve_link	Resolve a slug to its destination
analytics	Aggregate analytics across all your links

Install into a client

mcp install prints a ready-to-paste config block for your MCP client (Claude Desktop by default):

lnkify mcp install

It carries your resolved server URL and — if you authenticate with an API key — an LNKIFY_API_KEY entry, producing config like:

{
  "mcpServers": {
    "lnkify": {
      "command": "lnkify",
      "args": ["mcp", "serve"],
      "env": {
        "LNKIFY_API_URL": "https://lnkify.example.com",
        "LNKIFY_API_KEY": "lf_live_YOUR_KEY"
      }
    }
  }
}

Pass --write to merge it directly into the Claude Desktop config file:

lnkify mcp install --write

Restart your agent afterward to pick up the new server.

Next: Examples