--- title: I Built a Multi-Agent AI System with Agent Protocols.... description: A developer shares their experience building a multi-agent AI system with a Coordinator, Research, Calculator, and Summary Agent, and how debugging it revealed... canonical: https://daily.dev/posts/i-built-a-multi-agent-ai-system-with-agent-protocols-debugging-it-was-harder-than-building-it--qvvabn2zh twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: I Built a Multi-Agent AI System with Agent Protocols. Debugging It Was Harder Than Building It. | daily.dev og:description: A developer shares their experience building a multi-agent AI system with a Coordinator, Research, Calculator, and Summary Agent, and how debugging it revealed... og:url: https://daily.dev/posts/i-built-a-multi-agent-ai-system-with-agent-protocols-debugging-it-was-harder-than-building-it--qvvabn2zh og:image: https://api.daily.dev/og/posts/QVVABN2ZH.png og:image:alt: I Built a Multi-Agent AI System with Agent Protocols. Debugging It Was Harder Than Building It. og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # I Built a Multi-Agent AI System with Agent Protocols. Debugging It Was Harder Than Building It. **[Medium](https://daily.dev/sources/medium_js)** · 6 min read · 1 upvotes · 0 comments ## Summary A developer shares their experience building a multi-agent AI system with a Coordinator, Research, Calculator, and Summary Agent, and how debugging it revealed the real challenge: visibility. After struggling with unstructured print statements across concurrent requests, they instrumented the system with OpenTelemetry and SigNoz. Distributed tracing revealed that the Research Agent (waiting on an external service) was the true latency bottleneck — not the LLM — and exposed a race condition where the Coordinator continued before the Research Agent finished. Key lessons: structured inter-agent messaging with unique request IDs is essential, observability should be added from day one, and distributed tracing beats print-statement debugging at scale. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How can I debug a multi-agent AI system where print statements aren't enough? Instrument each agent with OpenTelemetry spans and visualize the traces with a tool like SigNoz. Wrapping each agent's execution in a tracer span (e.g. tracer.start_as_current_span) creates a unified timeline per request, correlating logs, timings, and dependencies so you can see exactly which agent failed or was slow instead of manually reading disconnected log lines. _daily.dev surfaces tracing and observability practices for engineers wiring up multi-agent pipelines._ ### Why is my multi-agent pipeline slow even though I assumed the LLM was the bottleneck? The slowest component may not be the LLM at all but a downstream agent waiting on an external service call. In one traced multi-agent request, the Research Agent took 1.6 seconds versus 30ms for the Coordinator, 90ms for the Calculator, and 420ms for the Summary Agent, because it waited on an external search service rather than model inference. _daily.dev helps developers tracking down real latency sources in agent-based architectures stay informed._ ### Why does my summary agent sometimes receive incomplete data in a multi-agent orchestration? This typically happens when the coordinating agent proceeds to the next step before an upstream agent, such as a research agent, has finished its work, rather than being a model quality issue. Distributed tracing can reveal that the coordinator continued early; fixing the orchestration logic to wait for all required agent responses resolves the incomplete output. _daily.dev keeps builders of agent orchestration logic current on debugging patterns like this._ ## Similar posts on daily.dev - [AI Agent Observability: Logging, Tracing, and Debugging Explained](https://daily.dev/posts/ai-agent-observability-logging-tracing-and-debugging-explained-95izimukh) · Machine Learning Mastery · 1 upvotes · 0 comments - [Who's monitoring the agents?](https://daily.dev/posts/who-s-monitoring-the-agents--lql2w0xc9) · The New Stack · 0 upvotes · 0 comments - [What Breaks When Multi-Agent Systems Scale](https://daily.dev/posts/what-breaks-when-multi-agent-systems-scale-y4xkecht3) · DigitalOcean Community · 1 upvotes · 0 comments - [Multi-agent observability: why one trace isn't enough](https://daily.dev/posts/multi-agent-observability-why-one-trace-isn-t-enough-yoykwfunh) · Redis · 0 upvotes · 0 comments - [AI Agent Observability 2026: Debug Production Agents](https://daily.dev/posts/ai-agent-observability-2026-debug-production-agents-sxo8wyw1v) · Alex CloudStar · 0 upvotes · 0 comments --- Tags: [#devops](https://daily.dev/tags/devops), [#ai-agents](https://daily.dev/tags/ai-agents), [#observability](https://daily.dev/tags/observability), [#opentelemetry](https://daily.dev/tags/opentelemetry) [View this post on daily.dev](https://daily.dev/posts/i-built-a-multi-agent-ai-system-with-agent-protocols-debugging-it-was-harder-than-building-it--qvvabn2zh) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"I Built a Multi-Agent AI System with Agent Protocols. Debugging It Was Harder Than Building It.","url":"https://daily.dev/posts/i-built-a-multi-agent-ai-system-with-agent-protocols-debugging-it-was-harder-than-building-it--qvvabn2zh","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/i-built-a-multi-agent-ai-system-with-agent-protocols-debugging-it-was-harder-than-building-it--qvvabn2zh"},"datePublished":"2026-07-21T16:07:27.128Z","dateModified":"2026-09-14T06:49:37.860Z","description":"A developer shares their experience building a multi-agent AI system with a Coordinator, Research, Calculator, and Summary Agent, and how debugging it revealed...","image":"https://media.daily.dev/image/upload/s--foaA6JGU--/f_auto/v1722860399/public/Placeholder%2004","thumbnailUrl":"https://media.daily.dev/image/upload/s--foaA6JGU--/f_auto/v1722860399/public/Placeholder%2004","isAccessibleForFree":true,"articleSection":"Medium","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Medium","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/medium","url":"https://daily.dev/sources/medium_js"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/i-built-a-multi-agent-ai-system-with-agent-protocols-debugging-it-was-harder-than-building-it--qvvabn2zh","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"devops,ai-agents,observability,opentelemetry","timeRequired":"PT6M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Medium","item":"https://daily.dev/sources/medium_js"},{"@type":"ListItem","position":3,"name":"I Built a Multi-Agent AI System with Agent Protocols. Debugging It Was Harder Than Building It."}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/i-built-a-multi-agent-ai-system-with-agent-protocols-debugging-it-was-harder-than-building-it--qvvabn2zh#faq","mainEntity":[{"@type":"Question","name":"How can I debug a multi-agent AI system where print statements aren't enough?","acceptedAnswer":{"@type":"Answer","text":"Instrument each agent with OpenTelemetry spans and visualize the traces with a tool like SigNoz. Wrapping each agent's execution in a tracer span (e.g. tracer.start_as_current_span) creates a unified timeline per request, correlating logs, timings, and dependencies so you can see exactly which agent failed or was slow instead of manually reading disconnected log lines. daily.dev surfaces tracing and observability practices for engineers wiring up multi-agent pipelines."}},{"@type":"Question","name":"Why is my multi-agent pipeline slow even though I assumed the LLM was the bottleneck?","acceptedAnswer":{"@type":"Answer","text":"The slowest component may not be the LLM at all but a downstream agent waiting on an external service call. In one traced multi-agent request, the Research Agent took 1.6 seconds versus 30ms for the Coordinator, 90ms for the Calculator, and 420ms for the Summary Agent, because it waited on an external search service rather than model inference. daily.dev helps developers tracking down real latency sources in agent-based architectures stay informed."}},{"@type":"Question","name":"Why does my summary agent sometimes receive incomplete data in a multi-agent orchestration?","acceptedAnswer":{"@type":"Answer","text":"This typically happens when the coordinating agent proceeds to the next step before an upstream agent, such as a research agent, has finished its work, rather than being a model quality issue. Distributed tracing can reveal that the coordinator continued early; fixing the orchestration logic to wait for all required agent responses resolves the incomplete output. daily.dev keeps builders of agent orchestration logic current on debugging patterns like this."}}]} ```