Docs
Home

Your own MCP servers

Add private MCP servers that only your agents can use, with headers or OAuth.

You can connect your own MCP servers — a personal automation, a Zapier endpoint, a self-hosted tool — without waiting for an admin. These servers are private to you: only agents acting on your behalf can use them.

Find them under Settings → My Connections.

Adding a server

  1. Click Add server.
  2. Fill in a display name, a URL, and an optional description. A short slug is suggested automatically (you can edit it).
  3. Choose how the server authenticates:
    • Headers — paste static headers (an API key or bearer token). Good for servers with a secret in a header, or a secret already baked into the URL.
    • OAuth — connect with a sign-in flow. You don't enter any client details: Monad sets everything up from the URL for you.
  4. Save.

Headers

Add one row per header (name + value). They're sent with every request to your server and stored encrypted.

OAuth

There's nothing to fill in. When you pick OAuth and save, Monad reads the server's sign-in details straight from its URL and registers with it for you — no client ID, secret, or sign-in addresses to copy from the provider.

After saving, open the server and click Connect to sign in. A green Connected badge means your agents can use it. Reconnect needed means you should sign in again.

Some servers don't support automatic setup yet. If yours doesn't, saving will tell you so. When that happens, if the server accepts an API key or token instead, add it using Headers rather than OAuth.

Managing servers

Each server in the list opens its own page (click the row). From there you can:

  • Enable / Disable — turn a server off without deleting it. The same toggle is on the list, so you can flip it without opening the server.
  • Edit — change any detail. For OAuth, there's nothing to re-enter; if you change the URL, Monad sets the connection up again automatically.
  • Test connection — check the server is reachable and see the tools, resources, and prompts it exposes.
  • Disable individual tools — after a test, switch off any tool you don't want your agents to use. Disabled tools are hidden from agents on their next run; tools the server adds later stay on by default.
  • Delete — removes the server and any stored credentials for good.

Renaming a slug: if you rename a server's slug, the old name is held for 24 hours before it can be reused, so anything still referring to the old name doesn't accidentally point at a different server.

Descriptions write themselves

The description is what your agents read when deciding whether to reach for a server, so keeping it accurate matters.

  • Leave it blank and Monad writes a short "what this is for" summary from the server's tools — and keeps it current, updating it shortly after the server's tools change so a server that gains new abilities is described correctly.
  • Write or edit it yourself and it's yours: Monad stops auto-updating it and never overwrites your text. Clear the field again if you'd like Monad to take over keeping it in sync.

Privacy

Your servers and their credentials are visible only to you. Admins and other members can't see or use them.