Wake, Notify, and Steer Long-Running Agents with Signals

​Long-running agents need a reliable way to receive new instructions and context after the initial request. A user might send a follow-up while the agent is reasoning, a background task might complete, or an external system might need to wake an idle thread. Signals make these interactions part of the agent’s ongoing loop.

​In this workshop, we’ll build an agent that can receive messages and system context in real time. We’ll explore how to wake idle threads, steer active runs, preserve turn order, and model durable state and external notifications.

​We’ll cover:

• ​Sending immediate and queued input with sendMessage(), queueMessage(), and queueSignal()

• ​Using sendSignal() to notify or steer an agent while it is running

• ​Representing changing context and external events with state and notification signals

• ​Subscribing to threads and coordinating long-running agents across distributed deployments

​Join us for a live demo with real code and the chance to ask questions directly to the engineers who built the feature.

​Hosted by

• ​Abhi Aiyer, Co-Founder and CTO, Mastra

  • ​/abhiaiyer

  • ​/abhi-aiyer-aa41bb42

• ​Tyler Barnes, Engineering, Mastra

  • ​/tylbar

  • ​/tyler-barnes-882a61193

​Recording and code examples will be available to everyone who registers.