Reported News
DeepSeek open-sources dsh-superagent, a prototype agent that runs GitHub requests in parallel sessions
dsh-superagent is an MIT-licensed coding agent DeepSeek put on GitHub on Oct 31, 2026: a DeepSeek Harness bundle that runs each allowed GitHub or Feishu request in its own agent session. Its README calls it a "Research prototype for reference only, not an official DeepSeek product."
- Changes reported
- 6
- listed below, each from a source we read
- We tested
- 0
- we did not run it ourselves
- Gaps we flag
- 4
- things we could not verify
- Sources
- 4
- 4 primary, linked below
What changed
| # | What the source says | Status |
|---|---|---|
| 01 | dsh-superagent is a proactive coding agent for a single GitHub repository, packaged as a bundle for DeepSeek Harness. It listens to GitHub and, optionally, Feishu, and answers in the original conversation, for example with a pull request. | Reported |
| 02 | It splits work between a router and workers: a main session decides whether a request should become a work item (a Ticket), and each accepted Ticket runs in its own worker session and working directory, with a configurable concurrency limit. | Reported |
| 03 | It acts only when an allowed person addresses it directly or when a rule the team wrote allows it; proactive work starts in a mode that only records what the agent would have done. | Reported |
| 04 | A web panel inside DeepSeek Harness lists first what needs a person (questions, reviews, blocked or failed work), then system health and the history of each request. | Reported |
| 05 | The deployment guide says both of its presets run with full access as the operator's OS user and that its account wrappers are not a security boundary; it recommends a dedicated OS account and a separate GitHub user for the agent. | Reported |
| 06 | It requires Node.js 22.19+ or 24+, GitHub CLI 2.48 or later and a repository on github.com (GitHub Enterprise Server is not supported), plus a model provider configured for DeepSeek Harness. | Reported |
reported by the sourcemeasured or run by usnot verified
Primary source: dsh-superagent repository and README (GitHub). We are the fast, independent index; the vendor’s own page is the source of record for this release.
In the source’s words
“Research prototype for reference only, not an official DeepSeek product.”
“take as little human attention as possible, and give as much value as possible.”
Early verdict
Worth reading for anyone designing team-facing coding agents, mainly for its permission and attention rules; treat it as research code, as DeepSeek itself does, and isolate it before running it on a real repository.
Basis: Based only on the repository's README and docs as read on Nov 1, 2026; we have not run it.
What we have not verified
- We did not install or run dsh-superagent or DeepSeek Harness; everything here comes from the repository's README and docs as read on Nov 1, 2026.
- Press coverage cites a DeepSeek simulation in which p95 task completion time fell from 29.5 to 15.7 minutes; that figure is not in the repository and we have not seen DeepSeek's own write-up, so we do not report it as fact.
- The README says the published code differs from DeepSeek's internal version and that some designs in its accompanying article are not implemented in this prototype; we did not check which.
- The source time is when the GitHub repository was created; we first saw the release through a press feed about a day later.
This brief is Reported: written from the sources below, not from running the release ourselves.
Sources
- dsh-superagent repository and README (GitHub)primary source · fetched 2026-11-01
- dsh-superagent deployment guide (docs/006-deployment.md)primary source · fetched 2026-11-01
- dsh-superagent philosophy (docs/000-philosophy.md)primary source · fetched 2026-11-01
- dsh-superagent execution model (docs/003-execution.md)primary source · fetched 2026-11-01
Related
Spotted an error? See our corrections policy or email the address on the contact page. Markdown version for agents: /news/deepseek-dsh-superagent-coding-agent-prototype.md
Want these briefs in your Google Top Stories? Add Agentic Weekly as a preferred source on Google.
The Weekly Verdict · free
5 new-model use cases. Every Monday.
A free email every Monday: the week's few big AI releases, then 5 short use cases for the new models, each with a prompt you can paste.
Free. Double opt-in. One click to unsubscribe.