You build something in Claude Code or Cursor. It produces a URL — a deploy, a file, a share link. Now you want a short link to put in a commit message, a tweet, or a README.
The old way: copy the long URL, open the shortener, paste it, copy the short link back. That is four round trips between your agent and your browser. After the third time, it gets old.
If you work in an AI agent, you can just ask it to shorten the URL. No browser, no paste, no signup.
The agent-first way
KooLink ships a hosted MCP server. If your agent speaks MCP — Claude Desktop, Claude Code, Cursor, Windsurf, or any of the growing set of MCP-native tools — you point it at the server and it gets a shorten_url tool.
Connection string:
https://mcp.kool.ink
No API key. No account. No free-tier cap. You point your agent at the server and it can shorten any URL in one call.
What it looks like in practice
Add the server to your MCP config:
{
"mcpServers": {
"koolink": {
"url": "https://mcp.kool.ink",
"type": "streamable-http"
}
}
}
Then ask the agent: "shorten https://example.com/some/very/long/path". It calls the tool and hands you back a compact kool.ink/xyz link.
Why one call matters
The friction is not the shortening. The friction is switching context. Every time you leave the agent to do a browser chore, you break flow. An MCP server turns that chore into a tool inside the same context where you are already working.
For people who spend their day inside an agent, the shortener should be there — not somewhere you go. That is the whole idea.
There is also a terminal command
Prefer your shell? The same service is one line away:
npx @koolink/koolink https://example.com/very/long/path
# → https://kool.ink/xyz
Done
Point your agent at https://mcp.kool.ink, or run npx @koolink/koolink. Either way you get a clean link without leaving the tool you are already in.
Free forever. Unlimited. No signup.