--- title: NEW DeepSeek Harness -  Super Agent! | daily.dev description: DeepSeek launched Super Agent, a prototype that watches GitHub issues and pull requests and autonomously completes tasks with a PR. It's built entirely from... canonical: https://daily.dev/posts/new-deepseek-harness---super-agent--p1xwixb2f twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: NEW DeepSeek Harness -  Super Agent! | daily.dev og:description: DeepSeek launched Super Agent, a prototype that watches GitHub issues and pull requests and autonomously completes tasks with a PR. It's built entirely from... og:url: https://daily.dev/posts/new-deepseek-harness---super-agent--p1xwixb2f og:image: https://api.daily.dev/og/posts/p1XwIXB2F.png og:image:alt: NEW DeepSeek Harness -  Super 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. # NEW DeepSeek Harness -  Super Agent! **[Sam Witteveen](https://daily.dev/sources/samwitteveenai)** · 17 min read · 0 upvotes · 0 comments ## Summary DeepSeek launched Super Agent, a prototype that watches GitHub issues and pull requests and autonomously completes tasks with a PR. It's built entirely from plugins on top of DeepSeek Harness (DSH), released August 13 alongside DeepSeek V4 Pro and MIT licensed. In a benchmark called Frontier Harness Eval, running the same model through different harnesses showed DSH and Claude Code both hit a 63.3% pass rate, but DSH cost about $328 per successful task versus $18.34 for Claude Code. Across all configurations tested, the cost gap between cheapest and most expensive was about 17.5x. Under stress tests, DeepSeek pushed trigger rates from 1 to 40 per minute while decision time per task dropped from 6.4 to 1.6 seconds through batching. The team admits prompt injection, credential leaking, and cross-device permissions remain unsolved. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How does DeepSeek Harness compare in cost to Claude Code for the same coding tasks? In the Frontier Harness Eval benchmark, DeepSeek Harness (in creator configuration) and Claude Code both achieved a 63.3% pass rate on the same 30 tasks using the same model, Kimmy Kade 3. However, DeepSeek Harness cost about $3.28 per successful task compared to $18.34 for Claude Code, roughly a 5.6x cost difference despite identical pass rates. _Comparing agent harness costs before picking one? Track benchmarks like this on daily.dev._ ### What is DeepSeek's Super Agent and how does it work on GitHub? Super Agent is a research prototype built on DeepSeek Harness that watches a GitHub repo's issues and pull requests. When an authorized team member requests something, it spins up an isolated worker session, completes the work, and replies with a pull request. It uses a two-layer architecture: a long-running 'router' session handling human interaction, and isolated worker agents that report back to the router rather than talking to people directly. _Exploring autonomous coding agent designs? Follow prototypes like this one on daily.dev._ ### What security risks remain unsolved in DeepSeek's Super Agent prototype? DeepSeek acknowledges that prompt injection, credential leaking, and permissions management across devices are problems it has not fully solved in the Super Agent prototype. Because of this, caution is advised when deploying it, though running it with local models in a fully locked-down environment reduces the risk. _Weighing the security tradeoffs of autonomous coding agents? Track these caveats on daily.dev._ --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#architecture](https://daily.dev/tags/architecture), [#claude-code](https://daily.dev/tags/claude-code), [#deepseek](https://daily.dev/tags/deepseek) [View this post on daily.dev](https://daily.dev/posts/new-deepseek-harness---super-agent--p1xwixb2f) ```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":"NEW DeepSeek Harness -  Super Agent!","url":"https://daily.dev/posts/new-deepseek-harness---super-agent--p1xwixb2f","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/new-deepseek-harness---super-agent--p1xwixb2f"},"datePublished":"2026-11-03T14:52:52.907Z","dateModified":"2026-11-03T14:54:02.432Z","description":"DeepSeek launched Super Agent, a prototype that watches GitHub issues and pull requests and autonomously completes tasks with a PR. It's built entirely from...","image":"https://i.ytimg.com/vi/OrcsMi2-IcM/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/OrcsMi2-IcM/sddefault.jpg","isAccessibleForFree":true,"articleSection":"Sam Witteveen","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":"Sam Witteveen","logo":"https://media.daily.dev/image/upload/s--gJm-KsgL--/f_auto/v1711727006/logos/samwitteveenai","url":"https://daily.dev/sources/samwitteveenai"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/new-deepseek-harness---super-agent--p1xwixb2f","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,architecture,claude-code,deepseek","timeRequired":"PT17M","video":{"@type":"VideoObject","name":"NEW DeepSeek Harness -  Super Agent!","description":"DeepSeek launched Super Agent, a prototype that watches GitHub issues and pull requests and autonomously completes tasks with a PR. It's built entirely from...","thumbnailUrl":"https://i.ytimg.com/vi/OrcsMi2-IcM/sddefault.jpg","uploadDate":"2026-11-03T14:52:52.907Z","duration":"PT17M","url":"https://api.daily.dev/r/p1XwIXB2F","embedUrl":"https://www.youtube.com/embed/OrcsMi2-IcM"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Sam Witteveen","item":"https://daily.dev/sources/samwitteveenai"},{"@type":"ListItem","position":3,"name":"NEW DeepSeek Harness -  Super Agent!"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/new-deepseek-harness---super-agent--p1xwixb2f#faq","mainEntity":[{"@type":"Question","name":"How does DeepSeek Harness compare in cost to Claude Code for the same coding tasks?","acceptedAnswer":{"@type":"Answer","text":"In the Frontier Harness Eval benchmark, DeepSeek Harness (in creator configuration) and Claude Code both achieved a 63.3% pass rate on the same 30 tasks using the same model, Kimmy Kade 3. However, DeepSeek Harness cost about $3.28 per successful task compared to $18.34 for Claude Code, roughly a 5.6x cost difference despite identical pass rates. Comparing agent harness costs before picking one? Track benchmarks like this on daily.dev."}},{"@type":"Question","name":"What is DeepSeek's Super Agent and how does it work on GitHub?","acceptedAnswer":{"@type":"Answer","text":"Super Agent is a research prototype built on DeepSeek Harness that watches a GitHub repo's issues and pull requests. When an authorized team member requests something, it spins up an isolated worker session, completes the work, and replies with a pull request. It uses a two-layer architecture: a long-running 'router' session handling human interaction, and isolated worker agents that report back to the router rather than talking to people directly. Exploring autonomous coding agent designs? Follow prototypes like this one on daily.dev."}},{"@type":"Question","name":"What security risks remain unsolved in DeepSeek's Super Agent prototype?","acceptedAnswer":{"@type":"Answer","text":"DeepSeek acknowledges that prompt injection, credential leaking, and permissions management across devices are problems it has not fully solved in the Super Agent prototype. Because of this, caution is advised when deploying it, though running it with local models in a fully locked-down environment reduces the risk. Weighing the security tradeoffs of autonomous coding agents? Track these caveats on daily.dev."}}]} ```