Home › Glossary › Systems › Agent Delegation

Intermediate · Systems

Agent Delegation

Visual diagram · (in preparation) · Math · (in preparation) · Worked example · 3 difficulty levels.

TL;DR. Handing part of a task, and the authority needed to complete it, from a human to an agent or from one agent to another.

Technical Definition

Handing part of a task, and the authority needed to complete it, from a human to an agent or from one agent to another.

How it works

Delegation is what makes multi-agent systems tractable: a planner decomposes work and hands sub-tasks to specialists, each receiving a goal, a bounded context, and narrowed authority. Sound delegation specifies the objective and its success criteria, the resources and permissions granted, the escalation rule when the sub-agent gets stuck, and the format of the result returned. Unbounded delegation is the usual root cause of runaway agent loops and surprise side effects.

Related Concepts

  • Agent Credential Delegation — Passing a scoped, verifiable credential from a principal to an agent — or from agent to sub-agent — so it can act on their behalf.
  • Delegated Authority — The specific, bounded permission a principal transfers to an agent so it can act on their behalf.
  • Agent-to-Agent Intent — A structured statement of purpose one agent sends another so the request can be understood, authorized, and negotiated.
  • Agent orchestration — Agent orchestration is the management of multiple AI agents or LLM calls to efficiently handle complex tasks.