Mastra introduces Signals to steer running agents
TypeScript AI framework Mastra has introduced Signals, a control mechanism designed to manage the lifecycle and execution of long-running AI agents. By enabling developers to send dynamic inputs, system instructions, and state updates directly to agents, Mastra allows human-in-the-loop steering and event-driven updates without restarts.
Dynamic execution steering via signals is the missing link for converting static agent scripts into persistent, interactive web services.
- –Control layer separation: Allowing external processes to send targeted signals means AI agents can now act as true background daemons that listen to live events.
- –Human-in-the-loop: Rather than stopping the entire workflow, agents can pause, await a signal containing user approval, and resume seamlessly.
- –Better state durability: Interacting with running agent threads without full execution context resets saves computational tokens and preserves memory state.
DISCOVERED
3d ago
2026-06-18
PUBLISHED
3d ago
2026-06-18
RELEVANCE
AUTHOR
mastra