mishal brands

dsh-superagent: developer tools for developers

Visit official site

Official site link, not an affiliate link. We do not have an approved referral link for dsh-superagent.

dsh-superagent helps you build software by starting with a research prototype that takes an allowed request on one GitHub repository and works on it in its own session. It is best for developers who are comfortable running research code on a machine kept separate from everyday work. People also use it to automate repetitive work. A free plan is available.

Illustrative preview: a generic illustration of the tool's job
Illustrative preview

What it does

dsh-superagent is a research prototype of a proactive coding agent for one GitHub repository. The README's principle is to take as little human attention as possible, and give as much value as possible. An allowed person asks, the agent works in its own session, and the reply goes back to the original conversation. The philosophy doc is docs/000-philosophy.md.

Repository: deepseek-ai/dsh-superagent-prototype. License: MIT. Setup: SETUP.md. Plain-English explainer: what it is, and who should wait. Open-source tools on this site: GitHub. Other pages in that section: agent-desktop and PromptDirector.

Editorial note

dsh-superagent is a research prototype from the DeepSeek Harness team. It watches one GitHub repository and, when an allowed person asks, works in its own session and replies in that same conversation. The README says it is not an official DeepSeek product. The code is MIT licensed.

Who it’s for

Key use cases

What you can do with it

How to use dsh-superagent: step-by-step tutorial

Five to eight steps from the account to a first useful result. Where the official guide names a control, the step uses that name. Where it does not, the step stays general.

  1. Read the README before you install anything. It says this is a research prototype for reference only, not an official DeepSeek product, and not production-ready software. Plan to run it on a computer, or an operating-system account, kept apart from everyday work.
  2. Install what the setup guide lists: Node.js 22.19 or newer, or Node.js 24 or newer; pnpm; git 2.32 or newer; Python 3.8 or newer; tmux; GitHub CLI 2.48 or newer; and DeepSeek Harness with a model provider configured. The deployment guide says GitHub Enterprise Server is not supported.
  3. Ask a coding agent to read SETUP.md in the repository and set dsh-superagent up with you. The file says the agent asks one question at a time, shows the full plan once, and installs only after you confirm.
  4. Give the agent its own GitHub account, with write access to one repository, and a GitHub CLI signed in only as that account. From an account on the allowed list, mention the agent on an issue. The setup guide says a work item should show on the panel within about a minute.
  5. Leave unasked work in the default observe mode until you have read what the agent would have accepted. The authorization doc says observe records a would-accept decision and does not start that work. A direct request from an allowed person still runs. Switch to auto only after the recorded decisions look right.

These steps follow official guides: README; SETUP.md; Deployment guide. Checked November 2, 2026.

Frequently asked questions

Is dsh-superagent free?

The code is. The license file is the MIT License, an ongoing free license, not a free trial. The deployment guide says you configure a model provider for DeepSeek Harness. The repository does not list a price for that model. Checked Nov 2, 2026.

Is it an official DeepSeek product?

No. The README says it is a research prototype for reference only, not an official DeepSeek product, and not production-ready software.

Who should try it?

Developers who are comfortable running research code on a machine kept separate from everyday work. Everyone else should wait.

Will it act without being asked?

Not by default. The authorization doc says observe is the default. Unasked work is recorded and not done until someone switches the accept mode to auto. A direct request from an allowed person still runs.

What machine does it need?

The setup guide lists Node.js 22.19 or newer, or Node.js 24 or newer; pnpm; git 2.32 or newer; Python 3.8 or newer; tmux; GitHub CLI 2.48 or newer; and DeepSeek Harness with a model provider configured.

Pricing

Prices checked on the official pricing page.

Open source

Free

MIT license for the code. The deployment guide also requires DeepSeek Harness and a configured model provider, which this repository does not price.

Facts

Access
Ongoing free plan
Paid plans from
Free
Checked
Category
Category: Development tools (31 in the library).
On the record
The license file is the MIT License, copyright 2026 DeepSeek Harness Team. This is an ongoing free license to the code, not a free trial and not a one-time credit grant. The deployment guide says you also configure a model provider for DeepSeek Harness. The repository does not list a price for that model.
Sources