Connect Wren to an AI client over MCP

Wren can connect to an outside AI client — Claude, ChatGPT, or Codex — so that assistant can drive your real Wren writing workflow: developing ideas, outlining, drafting, suggesting images, and scheduling posts against your actual account. Wren provides the substrate (your voice, audience, guardrails, and saved work); the connected AI does the thinking.

This connection uses MCP (Model Context Protocol), a standard way for AI clients to talk to tools like Wren.

The endpoint

Every connection points at the same Wren MCP endpoint:

https://app.writewithwren.com/api/mcp

You can copy this from Settings → Connected Accounts → MCP Server using the Copy button.

Two ways to connect

How you connect depends on your client:

• Claude and ChatGPT (and Codex desktop): add Wren as a custom connector and approve it in your browser. The client handles tokens for you — there's nothing to copy by hand.
• Codex CLI (or other headless setups): use a token you mint in Settings.

Connect a browser-based client

1. In your AI client, open its connector settings and choose to add a custom connector.
2. Paste the Wren endpoint above as the URL. (In ChatGPT only, set the Base scopes field to wren:read wren:write.)
3. Connect. A browser window opens to a "Connect to Wren" screen — make sure you're logged into Wren, then click Approve.

That's it. The client stores its own access automatically.

Mint a token (for CLI clients)

For clients that can't open a browser:

1. Go to Settings → Connected Accounts → MCP Server.
2. Click Create token, enter a label, pick a scope and an expiry (30 days, 90 days, or Never).
3. Your token is shown once — copy it right away and paste it into your client's configuration. You won't be able to see it again.

Each token appears in the list with its label, scope, last-used time, and status. To turn one off, click Revoke on its row; it stops working on its next use.

To disconnect a client you connected through your browser (Claude, ChatGPT), remove the Wren connector from that client's own settings — a browser-approved connection isn't a token in this list. The Revoke button here is for tokens you minted yourself (CLI clients).

Read-only vs. full scope

When you create a token (or connect a client), you choose a scope:

• Read-only — the AI can read your Wren content but not change anything.
• Full — the AI can read and also create or update ideas, posts, and scheduled work.

Pick read-only if you only want the AI to reference your account, and full if you want it to draft and save.

Publishing stays in your hands

Connecting an AI does not let it publish to LinkedIn on its own. There is no "publish now" action over MCP — posting immediately is always a manual step you take inside Wren. An AI can schedule a post for a future time, but only when you explicitly ask it to.

For deeper client-specific setup details, see Wren's MCP connector setup guide.