Multi-Agent System

Full definition

Common topologies include planner-worker (one agent decomposes, others execute), debater (agents argue to converge), supervisor-subordinate (hierarchical delegation), and swarm (peer-to-peer with shared scratchpad). Frameworks include Microsoft Agent Framework (successor to the now-archived AutoGen), CrewAI, LangGraph, and Anthropic's reference patterns. Multi-agent improves on single-agent for tasks with parallel sub-goals or where role specialization reduces context bloat.

Why it matters

Multi-agent is also where cost spirals and debugging gets hard: every agent adds latency, tokens, and failure modes. Use multi-agent when sub-tasks genuinely parallelize or specialize; use a single agent with tools otherwise.

Example

A research-report system uses one planner agent, four parallel researcher agents (markets, competitors, financials, regulatory), and one editor agent — producing in 12 minutes what previously took two analysts a day.

Next steps

• Get a free workflow map →We map one of your workflows and give you a written verdict: software, a process change or AI. No obligation.
• AI consulting cost guide →Hourly rates and project budgets by firm tier, from independents to the Big Four.

Related terms

• Agentic Workflow
  An agentic workflow is a multi-step AI process where a model plans, takes actions through tools, observes results, and iterates toward a goal — replacing single-shot prompting with a loop that reasons, acts, and self-corrects.
• AI Agent / Agentic AI
  An AI agent is an LLM-driven system that, given a goal, plans a sequence of steps, calls tools or APIs, observes the results, and iterates until the goal is reached — going beyond single-turn chat to multi-step autonomous action.
• ReAct Pattern
  ReAct (Reasoning + Acting) is a prompting pattern that interleaves reasoning steps ("Thought") with tool actions ("Action") and tool results ("Observation"), letting an LLM iteratively decompose a task, query the world, and self-correct.
• Tool Use (LLM)
  LLM tool use is the pattern in which a language model acts through external functions, APIs, databases or code execution: the model requests a tool, the host application runs it and returns the result, and the model continues, turning a text generator into a controller that can act. Function calling is the structured-output mechanism most tool use is built on.

More deployment & operations terms

• Dense Passage Retrieval (DPR)
  Dense Passage Retrieval (DPR) is a retrieval method that embeds queries and passages into the same vector space using two BERT encoders (a dual encoder) trained with contrastive loss on (question, positive-passage) pairs.
• Embedding Model
  An embedding model is a neural network that maps text, images, or other inputs into fixed-dimensional vectors where semantic similarity corresponds to geometric closeness (cosine or dot-product distance).
• Semantic Search
  Semantic search is information retrieval that ranks documents by meaning rather than exact-keyword overlap, using vector embeddings of query and documents to measure similarity in a learned semantic space.