Connect Claude, ChatGPT or Cursor

Connect Claude, ChatGPT, or Cursor to Quillly and any MCP-capable client can research, write, and publish straight to your domain. Quillly speaks the Model Context Protocol (MCP), so this page covers the one endpoint, the two ways to authenticate, and the exact steps to connect Claude, ChatGPT, and Cursor — then a few prompts to try first.

The MCP endpoint#

Every client connects to the same URL:

code
https://quillly.com/api/mcp

It is a stateless JSON-RPC 2.0 endpoint over HTTP (MCP protocol version 2025-06-18). You do not install anything — you paste this URL, or a small JSON snippet, into your client and approve it once.

Authentication#

Quillly accepts two credential types:

  • API key — created on the Integrations page, starts with qly_. Send it as a bearer token (Authorization: Bearer qly_...) or as a ?key=qly_... query parameter. Inside the product, onboarding can append ?key= to the URL for a one-paste setup with no OAuth.

  • OAuth 2.1 — for connectors that run an authorization flow (Claude, ChatGPT). An unauthenticated request returns 401, which triggers the client's OAuth flow; Quillly then issues a qat_ access token that grants every tool your plan allows.

Either way, your AI only ever sees the Quillly tools you enable — never your model provider's credentials. You can toggle individual tools on or off per API key on the Integrations page, and the same credential works everywhere a client runs — Claude on the web, desktop, and Claude Code all share one connector.

How an AI client connects to Quillly's MCP endpoint and publishes to your own domain

Connect Claude#

Requires Claude Pro or Max — custom connectors are a paid feature.

  1. Open claude.ai/customize/connectors.

  2. Click +Add custom connector → paste the URL above.

  3. Click Add and approve. Quillly appears in the + menu of every chat.

The same connector works in Claude Desktop and Claude Code.

Connect ChatGPT#

Requires ChatGPT Plus, Pro, Team, Enterprise, or Edu.

  1. Settings → Connectors → Advanced → enable Developer mode.

  2. Connectors → Add custom connector → paste the URL → authenticate.

  3. In any chat: + → Developer mode → turn on Quillly.

Connect Cursor#

Works on any Cursor plan. Add this to ~/.cursor/mcp.json, or a project-level .cursor/mcp.json:

json
{
  "mcpServers": {
    "quillly": {
      "url": "https://quillly.com/api/mcp"
    }
  }
}

Or use Settings (Cmd/Ctrl+Shift+J) → MCP → Add server, then approve Quillly — the tools load on the next agent run. Windsurf uses the same JSON with a serverUrl key in ~/.codeium/windsurf/mcp_config.json.

First prompts to try#

Once connected, talk to your AI in plain language:

  • "List my websites and show me what is published."

  • "Write a 900-word post on <topic>, add internal links, score it for SEO, and publish it to my blog."

  • "Draft a launch post for v2.0 from our CHANGELOG and publish it."

  • "Audit my lowest-scoring post and apply the SEO fixes."

New here? Start with Getting started with Quillly. To understand where blogs, docs, and changelogs live, see Content types & endpoints.