--- title: What Is an Agent Harness? The Architecture Behind Claude... description: An agent harness is defined as the runtime shell wrapped around an LLM that handles the loop calling the model, tool execution, memory, planning, and... canonical: https://daily.dev/posts/what-is-an-agent-harness-the-architecture-behind-claude-code-deepseek-harness-and-hermes-agent-yewiwnntj twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: What Is an Agent Harness? The Architecture Behind Claude Code, DeepSeek Harness, and Hermes Agent | daily.dev og:description: An agent harness is defined as the runtime shell wrapped around an LLM that handles the loop calling the model, tool execution, memory, planning, and... og:url: https://daily.dev/posts/what-is-an-agent-harness-the-architecture-behind-claude-code-deepseek-harness-and-hermes-agent-yewiwnntj og:image: https://api.daily.dev/og/posts/YeWiWnNTJ.png og:image:alt: What Is an Agent Harness? The Architecture Behind Claude Code, DeepSeek Harness, and Hermes Agent 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. # What Is an Agent Harness? The Architecture Behind Claude Code, DeepSeek Harness, and Hermes Agent **[freeCodeCamp](https://daily.dev/sources/freecodecamp)** · 23 min read · 12 upvotes · 1 comments ## Summary An agent harness is defined as the runtime shell wrapped around an LLM that handles the loop calling the model, tool execution, memory, planning, and sandboxing. Ten popular harnesses including Claude Code, DeepSeek Harness, Pi, and Hermes Agent are compared against a shared five-part architecture, revealing four competing engineering bets: DeepSeek's everything-is-a-plugin design, Claude Code and Deep Agents' fixed four-mechanism approach (planning, sandbox, subagents, context compression), Hermes Agent's compounding skill library, and Pi's radical minimalism. A 60-line Python implementation demonstrates the core agent loop, and the surrounding stack (MCP protocol layer, orchestration frameworks like LangGraph and CrewAI, and observability tools like Langfuse) is mapped out, alongside a decision guide for choosing a harness based on team needs rather than GitHub star counts. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What are the four core mechanisms that Claude Code's agent harness relies on, as identified by LangChain's Deep Agents project? LangChain's Deep Agents project identified four mechanisms behind Claude Code's harness: a planning tool that forces the model to write out steps before touching files, a virtual filesystem and sandbox for isolated repository access, subagent delegation where a smaller agent handles isolated work with its own clean context, and context/memory management middleware that compresses conversation history and offloads large tool outputs. _Engineers comparing agent harness designs can track breakdowns like this on daily.dev before picking an architecture._ ### What is the difference between an agent framework like LangChain and an agent harness like Claude Code? A framework such as LangChain, AutoGen, or CrewAI provides building blocks that developers import and wire together themselves, writing their own orchestration logic and choosing their own model calls. A harness like Claude Code ships the agent loop already built and opinionated about memory, planning, and safety, so the developer interacts with it by running a command rather than assembling components. _Developers weighing frameworks against harnesses can follow this distinction on daily.dev before committing a team's workflow._ ### How fast did DeepSeek's deepseek-harness grow on GitHub compared to Mario Zechner's Pi agent? Deepseek-harness passed 95,386 stars and 8,826 forks within two days of its August 13 launch, driven by a coordinated release and DeepSeek's existing audience. Pi, a solo-built four-tool coding agent, took roughly a year of organic word-of-mouth growth to reach a similar 91,600 stars with no single spike, suggesting steady retained usage rather than launch-driven attention. _Teams deciding which coding agent to adopt can weigh growth-curve signals like these on daily.dev._ ## Community discussion Top comments from developers on daily.dev. **@doncho** · 1 upvotes > I'm still trying to get oriented with all these new terms etc... ## Similar posts on daily.dev - [The Complete Guide to Agent Harnesses \(With Code\)](https://daily.dev/posts/the-complete-guide-to-agent-harnesses-with-code--wlssc2kw2) · Medium · 0 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#mcp](https://daily.dev/tags/mcp), [#claude-code](https://daily.dev/tags/claude-code), [#langchain](https://daily.dev/tags/langchain) [View this post on daily.dev](https://daily.dev/posts/what-is-an-agent-harness-the-architecture-behind-claude-code-deepseek-harness-and-hermes-agent-yewiwnntj) ```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":"What Is an Agent Harness? The Architecture Behind Claude Code, DeepSeek Harness, and Hermes Agent","url":"https://daily.dev/posts/what-is-an-agent-harness-the-architecture-behind-claude-code-deepseek-harness-and-hermes-agent-yewiwnntj","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/what-is-an-agent-harness-the-architecture-behind-claude-code-deepseek-harness-and-hermes-agent-yewiwnntj"},"datePublished":"2026-09-11T18:23:05.963Z","dateModified":"2026-10-15T02:22:35.382Z","description":"An agent harness is defined as the runtime shell wrapped around an LLM that handles the loop calling the model, tool execution, memory, planning, and...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/227aa2ff34776ab41a41c88f2b7f2876?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/227aa2ff34776ab41a41c88f2b7f2876?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"freeCodeCamp","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":"freeCodeCamp","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1628412854/logos/freecodecamp","url":"https://daily.dev/sources/freecodecamp"},"commentCount":1,"discussionUrl":"https://daily.dev/posts/what-is-an-agent-harness-the-architecture-behind-claude-code-deepseek-harness-and-hermes-agent-yewiwnntj","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":12},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":1}],"keywords":"ai-agents,mcp,claude-code,langchain","timeRequired":"PT23M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"freeCodeCamp","item":"https://daily.dev/sources/freecodecamp"},{"@type":"ListItem","position":3,"name":"What Is an Agent Harness? The Architecture Behind Claude Code, DeepSeek Harness, and Hermes Agent"}]} {"@context":"https://schema.org","@type":"WebPage","@id":"https://daily.dev/posts/what-is-an-agent-harness-the-architecture-behind-claude-code-deepseek-harness-and-hermes-agent-yewiwnntj","comment":[{"@type":"Comment","text":"I’m still trying to get oriented with all these new terms etc…","datePublished":"2026-09-13T15:08:43.461Z","url":"https://daily.dev/posts/YeWiWnNTJ#c-6yPmPTU6h","author":{"@type":"Person","name":"Doncho Angelov","url":"https://daily.dev/doncho","image":"https://media.daily.dev/image/upload/s--j1MSm0hy--/f_auto/v1716369534/avatars/avatar_EB0j0Umzh5v3zGWUSL5ma"},"interactionStatistic":{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1}}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/what-is-an-agent-harness-the-architecture-behind-claude-code-deepseek-harness-and-hermes-agent-yewiwnntj#faq","mainEntity":[{"@type":"Question","name":"What are the four core mechanisms that Claude Code's agent harness relies on, as identified by LangChain's Deep Agents project?","acceptedAnswer":{"@type":"Answer","text":"LangChain's Deep Agents project identified four mechanisms behind Claude Code's harness: a planning tool that forces the model to write out steps before touching files, a virtual filesystem and sandbox for isolated repository access, subagent delegation where a smaller agent handles isolated work with its own clean context, and context/memory management middleware that compresses conversation history and offloads large tool outputs. Engineers comparing agent harness designs can track breakdowns like this on daily.dev before picking an architecture."}},{"@type":"Question","name":"What is the difference between an agent framework like LangChain and an agent harness like Claude Code?","acceptedAnswer":{"@type":"Answer","text":"A framework such as LangChain, AutoGen, or CrewAI provides building blocks that developers import and wire together themselves, writing their own orchestration logic and choosing their own model calls. A harness like Claude Code ships the agent loop already built and opinionated about memory, planning, and safety, so the developer interacts with it by running a command rather than assembling components. Developers weighing frameworks against harnesses can follow this distinction on daily.dev before committing a team's workflow."}},{"@type":"Question","name":"How fast did DeepSeek's deepseek-harness grow on GitHub compared to Mario Zechner's Pi agent?","acceptedAnswer":{"@type":"Answer","text":"Deepseek-harness passed 95,386 stars and 8,826 forks within two days of its August 13 launch, driven by a coordinated release and DeepSeek's existing audience. Pi, a solo-built four-tool coding agent, took roughly a year of organic word-of-mouth growth to reach a similar 91,600 stars with no single spike, suggesting steady retained usage rather than launch-driven attention. Teams deciding which coding agent to adopt can weigh growth-curve signals like these on daily.dev."}}]} ```