--- title: LangSmith adds AI assistant and CLI tool for debugging... description: LangSmith has released two new tools aimed at debugging long-running, multi-step AI agents it calls 'deep agents.' Polly is an in-app AI assistant that lets... canonical: https://daily.dev/posts/langsmith-adds-ai-assistant-and-cli-tool-for-debugging-long-running-agents-7lkq5pmzv twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: LangSmith adds AI assistant and CLI tool for debugging long-running agents | daily.dev og:description: LangSmith has released two new tools aimed at debugging long-running, multi-step AI agents it calls 'deep agents.' Polly is an in-app AI assistant that lets... og:url: https://daily.dev/posts/langsmith-adds-ai-assistant-and-cli-tool-for-debugging-long-running-agents-7lkq5pmzv og:image: https://api.daily.dev/og/posts/7LkQ5PmZV.png og:image:alt: LangSmith adds AI assistant and CLI tool for debugging long-running agents 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. # LangSmith adds AI assistant and CLI tool for debugging long-running agents **[Collections](https://daily.dev/sources/collections)** · 2 min read · 1 upvotes · 0 comments ## Summary LangSmith has released two new tools aimed at debugging long-running, multi-step AI agents it calls 'deep agents.' Polly is an in-app AI assistant that lets developers chat with trace and thread data to spot inefficiencies, mistakes, or prompt improvement opportunities. langsmith-fetch is a CLI tool that connects coding agents like Claude Code or DeepAgents CLI to LangSmith traces, pulling data into local dev environments. The piece also notes the Warp terminal team is reportedly building a similar internal debugger that breaks down agent costs by task and subagent. ## Content Debugging agents is getting harder as they run longer. A simple LLM call produces a single trace you can scan in seconds. A deep agent - one that runs for dozens of turns, spawns subagents, and chains hundreds of tool calls - produces traces that are genuinely difficult to read, even for experienced engineers. LangSmith shipped two new features aimed at this problem. ## Polly: an in-app assistant for trace analysis Polly is a chat interface built into LangSmith that lets you ask questions about your trace and thread data directly. Instead of manually scanning hundreds of steps, you can ask where the agent went wrong, which subagent is driving up costs, or what prompt change might fix a recurring failure. It's meant to surface inefficiencies and suggest improvements without requiring you to read every node in the trace tree. ## langsmith-fetch: CLI access for local debugging The second tool is `langsmith-fetch`, a CLI that connects coding agents like Claude Code or DeepAgents to LangSmith traces and threads. The idea is to bring trace data into local workflows rather than forcing developers to context-switch into a web UI every time something breaks. ## Messages View: traces in conversation shape Separately, LangSmith also added a Messages View. The argument here is straightforward: for most builders, the fastest way to understand what went wrong is to replay the conversation and tool calls in the order the agent actually experienced them - not as a raw infrastructure trace. Messages View reformats the data into that shape, making it more accessible to developers who aren't primarily thinking about infra. ## Why this matters now The underlying issue is that deep agents are harder to debug than simple LLM calls for a few compounding reasons: prompts are longer, traces span more steps, and multi-turn interactions mean failures can originate several exchanges before they become visible. The cost dimension adds another layer - when subagents are running in parallel, it's not always obvious which task is responsible for a spike. Worth noting: Zach Lloyd at Warp is building something similar internally - a debugger that breaks down costs by agent and subagent task and shows agent processes over time. The problem is clearly on people's minds. ## Questions this post answers ### What new tools did LangSmith release for debugging long-running AI agents? LangSmith shipped two tools for debugging what it calls 'deep agents': Polly, an in-app AI assistant that lets developers chat with trace and thread data to spot inefficiencies and suggest prompt improvements, and langsmith-fetch, a CLI that connects coding agents like Claude Code or DeepAgents CLI to LangSmith traces so data can be pulled into a local environment. _Track new debugging tooling for multi-step agent workflows as it appears on daily.dev._ ### What is the difference between a run, a trace, and a thread in LangSmith's data model? A run is a single LLM call or tool invocation, a trace is the full tree of runs generated for one request, and a thread is a collection of multiple traces spanning a conversation with several turns. Deep agents stress all three because longer prompts add more to read, longer traces add more nesting, and multi-turn interactions require looking across threads. _Developers debugging multi-turn agent behavior can follow observability concepts like this on daily.dev._ ## Similar posts on daily.dev - [Debugging Deep Agents with LangSmith](https://daily.dev/posts/debugging-deep-agents-with-langsmith-v1oyf16pl) · LangChain · 1 upvotes · 0 comments - [Introducing Polly: Your AI Agent Engineer](https://daily.dev/posts/introducing-polly-your-ai-agent-engineer-knoawipym) · LangChain · 1 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#observability](https://daily.dev/tags/observability), [#claude-code](https://daily.dev/tags/claude-code), [#langsmith](https://daily.dev/tags/langsmith) [View this post on daily.dev](https://daily.dev/posts/langsmith-adds-ai-assistant-and-cli-tool-for-debugging-long-running-agents-7lkq5pmzv) ```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":"LangSmith adds AI assistant and CLI tool for debugging long-running agents","url":"https://daily.dev/posts/langsmith-adds-ai-assistant-and-cli-tool-for-debugging-long-running-agents-7lkq5pmzv","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/langsmith-adds-ai-assistant-and-cli-tool-for-debugging-long-running-agents-7lkq5pmzv"},"datePublished":"2026-08-27T18:32:51.362Z","dateModified":"2026-09-02T03:09:06.776Z","description":"LangSmith has released two new tools aimed at debugging long-running, multi-step AI agents it calls 'deep agents.' Polly is an in-app AI assistant that lets...","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/langsmith-adds-ai-assistant-and-cli-tool-for-debugging-long-running-agents-7lkq5pmzv","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,observability,claude-code,langsmith","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":"LangSmith adds AI assistant and CLI tool for debugging long-running agents"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/langsmith-adds-ai-assistant-and-cli-tool-for-debugging-long-running-agents-7lkq5pmzv#faq","mainEntity":[{"@type":"Question","name":"What new tools did LangSmith release for debugging long-running AI agents?","acceptedAnswer":{"@type":"Answer","text":"LangSmith shipped two tools for debugging what it calls 'deep agents': Polly, an in-app AI assistant that lets developers chat with trace and thread data to spot inefficiencies and suggest prompt improvements, and langsmith-fetch, a CLI that connects coding agents like Claude Code or DeepAgents CLI to LangSmith traces so data can be pulled into a local environment. Track new debugging tooling for multi-step agent workflows as it appears on daily.dev."}},{"@type":"Question","name":"What is the difference between a run, a trace, and a thread in LangSmith's data model?","acceptedAnswer":{"@type":"Answer","text":"A run is a single LLM call or tool invocation, a trace is the full tree of runs generated for one request, and a thread is a collection of multiple traces spanning a conversation with several turns. Deep agents stress all three because longer prompts add more to read, longer traces add more nesting, and multi-turn interactions require looking across threads. Developers debugging multi-turn agent behavior can follow observability concepts like this on daily.dev."}}]} ```