Google's Agent-to-Agent Protocol merges TaskPushNotificationConfig and PushNotificationConfig into a single type — reducing cognitive overhead and surface area for implementers building real-time agent systems.
PR #1500 consolidates two separate notification configuration types that served similar purposes:
TaskPushNotificationConfig — used when subscribing to task updatesPushNotificationConfig — used for general push notification configurationThese have been merged into a single PushNotificationConfig type. The breaking change marker (!) in the commit message signals that existing implementations will need to update their code.
Push notifications are how A2A enables real-time agent-to-agent communication. When Agent A delegates a task to Agent B, notifications allow B to push status updates back without A constantly polling.
Having two separate config types created friction:
This is part of a broader effort since the February 1.0 release. The A2A team is systematically identifying redundant or confusing constructs and simplifying them while the ecosystem is still young enough to absorb breaking changes.
The consolidated PushNotificationConfig now handles all push notification scenarios. The key fields remain:
{
"url": "https://agent-a.example.com/a2a/notifications",
"token": "bearer-token-for-auth",
"events": ["task.completed", "task.failed", "task.progress"]
}
The events field lets agents subscribe to specific notification types rather than receiving everything. This was previously only available in one of the two config types.
For implementers currently using TaskPushNotificationConfig:
TaskPushNotificationConfig with PushNotificationConfigThe same PR author (Herczyński) has a follow-up commit (010b9cc) removing config from binding entirely. The A2A team is clearly working through the spec systematically, and more consolidation PRs are likely incoming.
For agent developers, now is a good time to audit your A2A implementation against the latest spec — better to absorb these changes together than one at a time.