--- title: A local agent debugger that shares traces with Codex and... description: Lawrence Jones from incident.io shares how their team uses coding agents to debug their AI SRE product, which is itself a multi-agent system. Key techniques... canonical: https://daily.dev/posts/a-local-agent-debugger-that-shares-traces-with-codex-and-claude-code-unobgbhwg twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: A local agent debugger that shares traces with Codex and Claude Code | daily.dev og:description: Lawrence Jones from incident.io shares how their team uses coding agents to debug their AI SRE product, which is itself a multi-agent system. Key techniques... og:url: https://daily.dev/posts/a-local-agent-debugger-that-shares-traces-with-codex-and-claude-code-unobgbhwg og:image: https://api.daily.dev/og/posts/UnobgbHWg.png og:image:alt: A local agent debugger that shares traces with Codex and Claude Code 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. # A local agent debugger that shares traces with Codex and Claude Code **[Collections](https://daily.dev/sources/collections)** · 2 min read · 1 upvotes · 0 comments ## Summary Lawrence Jones from incident.io shares how their team uses coding agents to debug their AI SRE product, which is itself a multi-agent system. Key techniques include CLI-based eval tooling to avoid context limits, exporting trace data as a file system so agents can grep through it instead of using custom UIs, and parallel sub-agents to cluster failure patterns across thousands of daily backtests. The core insight: internal tooling for AI products must be designed to work well for AI agents — file systems and CLIs outperform custom interfaces and GUIs in agent debugging workflows. ## Content Lawrence Jones, a founding engineer at incident.io, wrote up how their team uses coding agents to debug and improve their AI SRE product — which is itself a complex multi-agent system. The core problem: when you have a large hierarchy of agents and prompts, figuring out *which* prompt is broken is genuinely hard. A few techniques they landed on: **CLI-based eval tooling.** They built a CLI tool that lets coding agents manage YAML-based prompt test suites directly. The reason: agents hit context limits fast when working through a UI, and a CLI keeps things scriptable and composable. **Exporting debug data as a file system.** Instead of building a custom UI or MCP integration for debugging, they export trace data as a downloadable file system. Claude Code can then just `grep` through it. This turns out to be far more useful than any purpose-built debug interface — agents are good at navigating file systems, and the approach scales without extra tooling overhead. **Parallel sub-agents for backtest analysis.** They run thousands of backtests daily. To make sense of failures at that volume, they use parallel sub-agents that cluster failure patterns and surface which prompts actually need fixing. Without this, the signal gets buried. The underlying insight is straightforward but easy to miss: if you're building internal tooling for a product that uses AI, that tooling also needs to work well *for* AI. File systems beat custom interfaces. CLIs beat GUIs. The debugging workflow has to be legible to the agents doing the debugging. Ben Hylak flagged something similar — that being able to share traces with Codex or Claude Code locally is the missing piece for agent debugging workflows. The incident.io approach is essentially a worked example of that idea in production. ## Similar posts on daily.dev - [How Databricks Uses AI to Accelerate Incident Investigation](https://daily.dev/posts/how-databricks-uses-ai-to-accelerate-incident-investigation-tyehhd3qg) · databricks · 5 upvotes · 2 comments - [AI DevOps vs. SRE agents: Compare AI incident response tools](https://daily.dev/posts/ai-devops-vs-sre-agents-compare-ai-incident-response-tools-mmosdnb8r) · The New Stack · 1 upvotes · 0 comments - [How we built an AI SRE agent that investigates like a team of engineers](https://daily.dev/posts/how-we-built-an-ai-sre-agent-that-investigates-like-a-team-of-engineers-w0ennf9ut) · Datadog · 0 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#observability](https://daily.dev/tags/observability), [#claude](https://daily.dev/tags/claude), [#codex](https://daily.dev/tags/codex) [View this post on daily.dev](https://daily.dev/posts/a-local-agent-debugger-that-shares-traces-with-codex-and-claude-code-unobgbhwg) ```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":"A local agent debugger that shares traces with Codex and Claude Code","url":"https://daily.dev/posts/a-local-agent-debugger-that-shares-traces-with-codex-and-claude-code-unobgbhwg","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/a-local-agent-debugger-that-shares-traces-with-codex-and-claude-code-unobgbhwg"},"datePublished":"2026-05-14T19:00:51.488Z","dateModified":"2026-05-17T16:57:45.314Z","description":"Lawrence Jones from incident.io shares how their team uses coding agents to debug their AI SRE product, which is itself a multi-agent system. Key techniques...","isAccessibleForFree":true,"articleSection":"Collections","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":"Collections","logo":"https://media.daily.dev/image/upload/s--fk_6ycEi--/f_auto,q_auto/v1780996001/logos/collections?_a=BAMAMiWQ0","url":"https://daily.dev/sources/collections"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/a-local-agent-debugger-that-shares-traces-with-codex-and-claude-code-unobgbhwg","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,observability,claude,codex","timeRequired":"PT2M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Collections","item":"https://daily.dev/sources/collections"},{"@type":"ListItem","position":3,"name":"A local agent debugger that shares traces with Codex and Claude Code"}]} ```