DigLabMCP DigLabDOCS
/
Get support
Connect your AI

Understand OAuth connections

Distinguish signing into an MCP connection from authorising its upstream services.

2 min read · Customer guide

OAuth is a supported authentication pattern, but each consent flow must be configured and tested for the exact destination. A successful API-key test does not establish OAuth readiness for a particular AI client.

An AI client may authenticate to an installation or its gateway using a supported first-party sign-in flow. Separately, a server may need OAuth credentials to call an upstream service on behalf of a business account.

These are not interchangeable. An upstream OAuth broker may need provider authorisation and token endpoints, client registration details and scopes. Do not switch an existing API-key connection to OAuth merely because the client offers an OAuth button.

Follow the agreed connection flow#

Use the exact server or gateway address in the handoff. When a browser consent screen opens, check the domain, account and requested access. Authorise only the intended audience and scope using the provider's normal process.

Do not send a callback URL containing an authorisation code to support or paste it into an AI conversation. Keep tokens and client secrets in the supported encrypted configuration route.

Confirm that the AI client returns to a connected state, discovers only the intended tools and completes the agreed read-only action in the correct account. Where refresh is part of the configured flow, verify it rather than assuming the initial connection will last indefinitely.

Record the stage and redacted error. A mismatched redirect address, unsupported registration flow, wrong account, missing scope or expired upstream grant can require different fixes. Do not disable authentication, accept a certificate warning or reuse another person's token as a workaround.

Updated 22 October 2026