Skip to main content

Gateway settings

Updated on 11 November 2026.

These settings apply to every agent and caller in your organization. Go to Settings > AI Gateway > A2A Gateway and open the Settings tab. Summary at the top shows what is in effect.

Gateway address​

Gateway address shows the URL every agent URL starts with: https://<gateway-host>/a2a/<tenant-handle>/<agent-slug>. The tenant handle identifies your organization; it is fixed. Copy the full URL of an agent from its Endpoints box.

Defaults​

SettingWhat it does
Default max hop depthThe deepest delegation chain allowed, 1 to 16 (default 5). An agent can set its own. Loop detection is always on.
Require signed agent cards by defaultAgents that do not set their own rule accept only cards with a verified signature.
Relay push notificationsAgents can send task updates to callers' push endpoints through the gateway, with data protection applied. Off removes push support from every agent card the gateway serves.
Public agent catalogAnyone can list the agents marked with a public card. Calls still need a caller key.

Click Save settings to apply changes.

Data protection​

The gateway scans requests, responses, streams and push updates with your organization's detection categories.

  • Requests to agents and Responses to callers each take an action: Allow, Flag, Redact or Block.
  • Category overrides set a different action for one category. Click Add override and pick the category and action.

A blocked call returns -31451. Redaction rewrites text and data values in place; file contents are never rewritten.

If a scan cannot finish (for example, it times out or a file type is not supported), the message goes through and the log records why.

Card signing keys​

The gateway signs the agent cards it serves, so callers can check that a card came from your gateway. The public keys are published at /a2a/<tenant-handle>/.well-known/jwks.json.

Click Rotate to start signing with a new key. The previous key stays published for 30 days, so cards signed with it keep verifying.

Who can read message content​

A2A logs store full message content. Who can read it is set by the A2A message content permission in your console roles. How long logs are kept follows your data retention settings.