Your AI agent does real work now. It writes code, opens pull requests, and answers questions. But somewhere in that workflow there is still one chore that pulls you back to the browser: creating a short link.
MCP changes that. If your agent speaks the Model Context Protocol, you hand it a shortener tool — and shortening becomes a call, not a trip across three tabs.
What MCP gives a shortener
The Model Context Protocol is how AI agents expose and consume tools. A shortener that offers an MCP server is not "a website with an API." It is a tool your agent already has loaded. When you ask the agent to shorten a long URL, it calls the tool and returns the short link — without leaving the context where you are already working.
The difference is not the shortening. It is the context switch. An MCP-integrated shortener keeps you inside the agent.
Connect to the KooLink MCP server
KooLink is a hosted MCP server. Point your client at it:
https://mcp.kool.ink
No API key. No account. No free-tier cap. It is a streamable-http server, so modern MCP clients connect directly.
In Claude Desktop or Code, it looks like this:
{
"mcpServers": {
"koolink": {
"url": "https://mcp.kool.ink",
"type": "streamable-http"
}
}
}
Once connected the agent gets a shorten_url tool. Ask it to shorten any URL and it does — one call, one clean kool.ink/xyz link back.
Or from the terminal
The same service is a one-liner away if you work in a shell:
npx @koolink/koolink https://example.com/very/long/path
# → https://kool.ink/xyz
The moat
Bitly does not expose an MCP server. Neither do TinyURL, Short.io, or Rebrandly as far as their current docs go. As agents become where a growing share of work actually happens, a shortener should be inside the agent — not a tab you switch to.
That is the entire point of KooLink being agent-first.
Do it
- Agent: point it at
https://mcp.kool.ink - Terminal:
npx @koolink/koolink <url> - Browser: head to kool.ink
Free forever. Unlimited. No signup.