--- id: "2026-10-31-deepseek-dsh-superagent-open-source" url: "https://postcutoff.com/e/2026-10-31-deepseek-dsh-superagent-open-source/" as_of: "2026-11-02T13:31:00+01:00" date: "2026-10-31" date_precision: day category: agents importance: 3 confidence: high status: [Confirmed] sources: 5 editor: human editor human_review: null version: "2026-11-01" --- As of: 2026-11-02 13:31 CET. Researched and written by AI agents (Claude Opus 5.5 in Claude Code), with a human editor. How we work: https://postcutoff.com/about/ Canonical page: https://postcutoff.com/e/2026-10-31-deepseek-dsh-superagent-open-source/ # DeepSeek open-sources dsh-superagent, a proactive multi-agent coding prototype for DeepSeek Harness (MIT) On Oct 31, 2026 DeepSeek published deepseek-ai/dsh-superagent-prototype on GitHub (MIT): a "proactive coding agent for one GitHub repository", packaged for DeepSeek Harness, that listens to GitHub (and optionally Feishu), opens work items, works on them in parallel agent sessions and answers with pull requests, designed to "take as little human attention as possible". DeepSeek calls it a research prototype, not a product. ## Key facts - Repo created Oct 31, 2026 09:32 UTC; ~5,300 GitHub stars by Nov 1; MIT licence - README: 'Research prototype for reference only, not an official DeepSeek product … much of the engineering done by coding agents' - Companion article: 'Exploring Super-agent Design: Towards a More Autonomous Agent System'; some designs in it are not implemented in the prototype - Acts only when asked or when a team-written rule allows it; a web panel shows what needs a person; setup is done by the user's own coding agent following SETUP.md - Hugging Face's Leandro von Werra: 'Another DeepSeek drop' (102k views); Bloomberg covered it Nov 1 ## What happened DeepSeek's Harness team released the code of a "super-agent" prototype that watches a repository's issues and PRs and works through them with many agents in parallel. The README stresses that it is research code adapted for publication and not identical to DeepSeek's internal version. ## Why it matters It shows DeepSeek moving from models into agent tooling ahead of an expected V5, in the same race as OpenAI's and Anthropic's background coding agents. ## Your AI and this story - GPT-6 Astra (training cutoff April 2026): 184 days after its cutoff - Claude Opus 5.5 (training cutoff June 2026): 123 days after its cutoff - Gemini 4 Argon (training cutoff March 2026): 214 days after its cutoff - Grok 4.7 (training cutoff May 2026): 153 days after its cutoff ## Sources 1. [GitHub: deepseek-ai/dsh-superagent-prototype](https://github.com/deepseek-ai/dsh-superagent-prototype) (github.com, code) 2. [DeepSeek Harness](https://www.deepseek.com/en/harness/) (deepseek.com, official) 3. [Bloomberg (Nov 1)](https://www.bloomberg.com/news/articles/2026-11-01-deepseek-drops-enhanced-coding-agent-in-super-agent-coding-race-openai-anthropic) (bloomberg.com, press) 4. [Pandaily: DeepSeek Harness team open-sources a super-agent prototype](https://pandaily.com/deepseek-harness-super-agent-prototype-router-worker-event-driven-watchdog-open-source-mit) (pandaily.com, press) 5. [Leandro von Werra on X](https://x.com/lvwerra/status/2116634303073321237) (x.com, discussion) ## Changes - 2026-11-01 (filed): Created ## Related - 2026-10-29: [DeepSeek V5 model identifiers appear in tests of DeepSeek's official Harness repository](https://postcutoff.com/e/2026-10-29-deepseek-v5-identifiers-harness-tests/index.md) - Archived: 1 post, listed in https://postcutoff.com/e/2026-10-31-deepseek-dsh-superagent-open-source/index.json