One MCP endpoint for ChatGPT, Claude, Cursor, plus a skill file so your agent knows what to do once it's connected.
Two minutes: add the server, connect your account, then just ask.
In your MCP client's settings (Claude, Cursor, ChatGPT, anything that speaks MCP), add a streamable-HTTP server with this URL:
https://linkversity.lol/mcp
{
"mcpServers": {
"linkversity": {
"type": "streamable-http",
"url": "https://linkversity.lol/mcp"
}
}
}
The first time you use a Linkversity tool, your client hits the server's sign-in challenge (HTTP 401) and opens the consent screen: approve once and it gets collections:read + collections:write. Your client fetches and refreshes the token itself (7-day access, 90-day refresh), so you never handle credentials. Public search stays open on the website, no account needed there.
Natural language is the whole interface:
“Find public collections about systems design, then save this article to my Backend collection.”
One SKILL.md: endpoint, all 7 tools with their scopes, the save-by-name workflow and the guardrails. Point your agent's skill loader at it, or drop the markdown into its skills folder as SKILL.md.
https://linkversity.lol/agents/skill
collections:read / collections:write it needssave_link_to_collection) vs by ID (save_link)/api/v1 with the 120 req/min limit