Skip to main content
TeamDuo is a remote MCP server. Claude alone has several distinct ways to add one, and which applies depends on which Claude product you’re using — a chat client, a team workspace, a terminal, or your own code calling the API directly. This page covers all of them.

The connector URL

That single URL is the whole configuration for every path below except one — the API, which also needs a credential, since there’s no browser to complete an OAuth flow in.

Connect Claude

These three share one setting, because a remote connector is brokered through your Claude account, not your device — the connection to TeamDuo runs from Anthropic’s servers, not your machine, even when you added it from the Desktop app.
1

Open Connectors settings

In claude.ai, Claude Desktop, or Cowork: Settings → Connectors.
2

Add a custom connector

Click Add custom connector, paste the URL above, and click Add.
3

Connect

Back in the connector list, click Connect and approve TeamDuo in the browser tab that opens. You’re authorizing as yourself — the agent will reach exactly what your TeamDuo account can reach, and nothing more.
This is a Claude account limit, not a TeamDuo one: Claude’s Free plan allows one custom connector; Pro and Max have no limit on personal connectors.
This is separate from Claude Desktop’s claude_desktop_config.json, which configures local stdio servers that run on your machine. TeamDuo is a remote HTTP server, so it’s added through Connectors settings above, never through that file.

Other MCP clients

Any client that supports remote MCP servers over OAuth can connect the same way: add the connector URL, and follow its own authorization flow. ChatGPT, for one, adds custom connectors under its own connector settings, the same shape as Claude’s.

Clients that need an API key

For anything that can’t complete an interactive OAuth flow — scripts, CI, a self-hosted agent — create an API key under Settings → API keys and send it as a bearer token:
Keys carry scopes, and a key is the only way to reach the connector without a browser session. Treat one as equivalent to your read access to every exposed datasource its scopes allow.

Transports

Streamable HTTP is the current standard and what Claude uses. SSE exists for older clients that haven’t moved off it yet.

Verifying the connection

Ask the agent to list your teams, then your datasources. If the teams come back but the datasource list is empty, the connection is fine and nothing has been exposed yet — see Connect a datasource.

Calling the same data over plain HTTP

MCP is the recommended path for an agent. For a script or service that would rather speak raw HTTP, the same read-only surface — list teams, list datasources, describe one, run a query — is available as a small REST API. See the REST API overview.