CCA-F: Claude Certified Architect — Foundations

By Anthropic · Question Mode

QuestionQ74

Agentic Architecture & Orchestration

You are building a multi-agent research system with the Claude Agent SDK. A coordinator agent delegates work to specialized subagents: one for web search, one for document analysis, one for synthesizing findings, and one for generating reports. The system researches topics and produces comprehensive, cited reports.

When the system analyzes complex legal cases that cite multiple precedents, the document-analysis subagent currently processes each precedent sequentially. For a landmark case citing 12 precedents, this takes over 3 minutes to complete.

What is the most effective way to reduce this latency while still preserving the coordinator's ability to monitor and debug the system?

Explanation

Because the 12 precedents can be analyzed independently of one another, this is a classic case for the orchestrator-worker (fan-out/fan-in) pattern: the coordinator itself spawns multiple parallel document-analysis subagents, splits the precedents among them, and then aggregates their outputs before handing off to the synthesis step. Keeping the coordinator as the entity that spawns and tracks the subagents (rather than letting a subagent spawn its own children, building a deep recursive hierarchy, or offloading task distribution to an external queue) preserves a single, centralized point of visibility into all active agents, which is essential for monitoring progress and debugging failures — while still achieving the latency reduction from parallel execution.

Community Discussion

No comments yet. Be the first to start the discussion!