# DeepSeek Super-agent — an open coding agent that takes work from GitHub chats

> DeepSeek's open prototype of an agent that watches a repo's conversations, takes approved requests and answers with finished work.

- Org: DeepSeek
- Date: 2026-10-31
- Category: repo, tool
- Source: https://github.com/deepseek-ai/dsh-superagent-prototype
- Canonical page: https://ai-tldr.dev/releases/deepseek-dsh-superagent/

## In short

DeepSeek open-sourced dsh-superagent, an MIT research prototype of a proactive coding agent for one GitHub repo. It runs on DeepSeek Harness, turns approved requests into tracked work and replies with results such as pull requests.

## What it is

dsh-superagent is DeepSeek's open-source prototype of a proactive coding agent for a single GitHub repository, published on 31 October 2026. When an allowed person asks it for work on GitHub or in Feishu, it opens a tracked work item, does the job in its own agent session and directory, and replies in the original conversation, for example with a pull request. DeepSeek calls it a research prototype from the team's daily work, not an official product.

## How it works

The system is a plugin for DeepSeek Harness. Source plugins turn GitHub and Feishu events into Triggers; a main router session judges each one and either ignores it or opens a Ticket with its own authority rules. Each Ticket runs in a separate worker session from a limited pool, state lives in one SQLite database, and replies go out through an outbox to the channel the request came from. A web panel lists what needs a person first.

## Why it matters

Most coding agents wait to be opened in a terminal or IDE; this design puts the agent inside a team's existing conversations and keeps human attention low, which DeepSeek states as its main goal. Proactive work needs a rule the team wrote, and it starts in a mode that only records what the agent would have done, so teams can watch before they trust it.

## Who it's for

teams running coding agents on a shared GitHub repo

## Try it

Ask a coding agent: "Read SETUP.md in this repository and set up dsh-superagent with me."

## Quick facts

- Maker: DeepSeek
- License: MIT
- Language: TypeScript
- Runs on: DeepSeek Harness (dsh) plugin bundle
- Sources: GitHub (required), Feishu (optional)
- Status: Research prototype, not production-ready

## FAQ

### Is DeepSeek Super-agent the same as DeepSeek's internal version?

No. The dsh-superagent README says the public code differs from DeepSeek's internal version and from the implementation described in the essay 'Exploring Super-agent Design: Towards a More Autonomous Agent System'. Some designs in that essay are future directions and are not implemented. DeepSeek warns the prototype may contain bugs and needs review and adaptation before real use.

### What do I need to run dsh-superagent?

dsh-superagent's SETUP.md lists Node.js 22.19+ or 24+, pnpm, git 2.32+, tmux, Python 3.8+, GitHub CLI 2.48+ and DeepSeek Harness with a configured model provider. It also needs a dedicated GitHub account for the agent, invited to the target repository. Feishu is optional and needs its own bot app. The panel runs on port 3082 by default.

### Can DeepSeek Super-agent act without being asked?

Only when a rule the team wrote allows it. DeepSeek's design notes say the agent acts when an allowed person addresses it directly, or when a team-written rule permits proactive work, and that proactive work starts in a mode that only records what the agent would have done. The default accept mode in SETUP.md is 'observe'.

### How does dsh-superagent relate to DeepSeek Harness?

DeepSeek Harness (dsh) hosts dsh-superagent as a plugin bundle. The harness provides the agent sessions, the event stream and the plugin manager, plus presets such as 'dsh-superagent-router' for the main session and a worker preset for executors. Setup creates a separate 'superagent' profile with the command 'dsh superagent --from-default-profile'.

## Links

- [GitHub Repo](https://github.com/deepseek-ai/dsh-superagent-prototype)
- [Setup Guide](https://github.com/deepseek-ai/dsh-superagent-prototype/blob/main/SETUP.md)
- [Architecture](https://github.com/deepseek-ai/dsh-superagent-prototype/blob/main/docs/007-architecture.md)
- [Design Philosophy](https://github.com/deepseek-ai/dsh-superagent-prototype/blob/main/docs/000-philosophy.md)
- [DeepSeek Harness](https://github.com/deepseek-ai/deepseek-harness)