MCP Connection Quickstart
Connect Claude, Cursor, ChatGPT, or any MCP-compatible agent to your BlogNow workspace. Paste one URL, click Approve in the browser, and your agent can research, draft, audit, and publish posts on your behalf — using your model, your prompts, your guardrails.
What You'll Achieve
- Add the BlogNow MCP server to your AI client in one line
- Approve the connection with a single click in your browser
- Run your first agent-driven content workflow
TL;DR
BlogNow speaks the MCP OAuth 2.1 spec with dynamic client registration. You don't create API keys manually — the AI client registers itself, you authorize it in a browser, and an access token is issued automatically.
All you give your agent is the URL: https://mcp.blognow.tech
Before you start
- An active BlogNow workspace (free trial is fine).
- An MCP-compatible client — Claude Desktop, Cursor, ChatGPT (with MCP connectors), Cline, or any agent built on the MCP SDK.
- Workspace admin role if your agent will need
writeaccess (publishing, drafting). Members can grantreadonly.
Add the BlogNow MCP server to your client
One URL, no API key. The client handles registration on first connect.
Open Settings → Connectors (or edit claude_desktop_config.json directly) and add:
{
"mcpServers": {
"blognow": {
"url": "https://mcp.blognow.tech"
}
}
}No headers, no Authorization, no env vars. Claude discovers the OAuth metadata at /.well-known/oauth-authorization-server and registers itself.
Open Settings → Features → Model Context Protocol and add a new server with the same URL:
Name: blognow
URL: https://mcp.blognow.techIn a custom GPT or Project: Add MCP connector → paste https://mcp.blognow.tech → save. ChatGPT will trigger the OAuth flow on first use.
The official SDK handles DCR + PKCE if you pass an OAuthClientProvider:
import { Client } from "@modelcontextprotocol/sdk/client/index.js";
import { StreamableHTTPClientTransport } from "@modelcontextprotocol/sdk/client/streamableHttp.js";
const transport = new StreamableHTTPClientTransport(
new URL("https://mcp.blognow.tech"),
{
authProvider: yourOAuthClientProvider,
}
);
const client = new Client({ name: "my-agent", version: "1.0.0" });
await client.connect(transport);Approve the connection in your browser
First time your agent talks to BlogNow, it triggers an OAuth flow. You stay in control.
The first request from your AI client opens a browser tab on blognow.tech/oauth/connect with these steps:
- Sign in (or your existing session is reused).
- See the connector card — your AI client's name (e.g. Claude) and the scopes it's asking for (
readorread write). - Pick the workspace the agent should access (you can connect the same AI client to multiple workspaces separately).
- Click Approve.
Your browser redirects back to the AI client with a short-lived code, which the client immediately exchanges for an access token. That token never touches your filesystem — it lives only in the AI client's secure storage.
Connecting on a headless server? The MCP SDK emits a URL when no browser is available — open it on any device signed into your BlogNow account, approve, and the token is delivered to your server over the redirect URI you registered.
Verify the tools are live
Confirm your agent can see the BlogNow toolset.
- In Claude Desktop or Cursor, open the MCP servers panel. blognow should show a green status dot.
- Hover the server to see the tool list. You should see ten tools, including
list_posts,search_posts,get_post,create_post,update_post,publish_post, andlist_categories/list_tags.
Server not connecting? Common causes: the browser tab was closed before Approve, your client's redirect URL got blocked by a corporate proxy, or your selected workspace's plan has expired. Re-run the connect flow from your client's server panel.
Run your first agent flow
Hand the agent a real task and watch the tools fire.
Paste this into Claude Desktop, Cursor, or ChatGPT to confirm the round-trip works end-to-end:
Using the blognow MCP server:
1. List the 5 most recent posts in my workspace.
2. Pick one as a reference and draft a follow-up post titled
"What changed in 2026" — 800 words, same brand voice, save as draft.
3. Show me the draft URL when you're done.The agent will call list_posts, get_post, and create_post in sequence, then return a link to the new draft in your dashboard.
What happened behind the scenes
The whole connect flow is standards-based — no BlogNow-specific SDK, no copy-pasting tokens. Here's the dance:
- Your AI client fetched
/.well-known/oauth-authorization-server(RFC 8414) to discover the endpoints. - It registered itself via
/oauth/register(RFC 7591 Dynamic Client Registration) — no manual API key setup on either side. - It generated a PKCE pair (RFC 7636) and sent you to
/oauth/authorizeto grant scopes. - After Approve, the browser came back with a one-time code. The client exchanged it at
/oauth/tokenusing the PKCE verifier — only the client that started the flow can complete it. - The returned access token is scoped to one workspace and one AI client. Revoking it in your dashboard disconnects only that client; everything else keeps working.
Managing your connections
Every approved AI client shows up in your dashboard under Settings → API Keys, tagged with the AI client's name (Claude, Cursor, etc.). From there you can:
- See when the connection was last used.
- Revoke an AI client's access without affecting others.
- Rotate the underlying token without re-approving.
Tokens auto-expire after extended inactivity. Your agent will prompt you to re-approve from inside the AI client — no manual re-issuing needed.
Next Steps
Tool reference
Full list of MCP tools your agent can call — inputs, outputs, and scopes.
Scopes & security
How read vs write scopes work, how to scope a connection down, and incident response.
Need Help?
Stuck on the connection? Send us the AI client name and the redirect URL — we usually reply within a few hours.