liveBook · Manning

Pass-off pattern

As explained in AI Agents in Action, Second Edition

A pass-off pattern explicitly controls the information and communication passed between agents during a multi-step flow.

The pass-off pattern explicitly controls what information and communication agents pass between one another during a multi-step flow. It suits complex behavior, tighter control, and communication with external agents. section 4.4.3

Use the pattern when you need to:

  • Control agent-to-agent communication rather than rely only on conversational flow behavior. section 4.4.3
  • Add guardrails, retries, and fallback output around intermediate agent results. section 4.4.3
  • Connect to external agents while preserving more deliberate control over the flow. section 4.4.3

Guarded intermediate handoff

A research agent produces a result, a thinking agent turns it into a research plan, and a filesystem agent consumes the final output. An output guardrail checks whether the plan is sufficiently detailed; failed checks can retry up to three times before returning a fallback message.

How AI Agents in Action, Second Edition covers it

How deep it goes
Contrasts pass-off with conversational handoff and orchestration, then demonstrates guardrails, retries, and fallback handling rather than full pass-off internals.
Examples
Uses research planning, typed output validation, retry loops, fallback messages, filesystem agents, and orchestration tools.
What you can do after
Apply it to design controlled agent flows with validated outputs, bounded retries, and explicit failure behavior.
How it connects
Links agent flows with handoff patterns, conversational flows, orchestration, guardrails, retries, and external agents.

This explainer was generated by Manning's knowledge graph from the text of AI Agents in Action, Second Edition. The section links point to the parts of the book it draws on.