# DeepSeek Harness — card for AI agents

An agent harness from DeepSeek in which every capability is a plugin. Its agent reads and edits workspace files, runs commands, delegates work and keeps a plan, and it runs as a desktop app, a browser UI or the dsh command-line tool.

Maker: DeepSeek · category: General-purpose AI agents · open source (MIT)

## Price
- Free and open source (MIT); users enter their own DeepSeek API key or another OpenAI-compatible provider.
- Free tier: yes
- Pricing page: https://deepseek-harness.github.io/deepseek-harness/en/guide/quickstart
- Checked: 2026-11-21

## Install
- npm: `npm install -g @deepseek-ai/dsh`

## Facts
- Interfaces: Desktop app, Web app, CLI, Self-hosted
- Repository: https://github.com/deepseek-ai/deepseek-harness

## Numbers (AgentGid data)
- Gid Score: 67 (rank #16) (public popularity and activity, not quality)
- Downloads last 7 days: 676,073
- GitHub stars: 284,792

## Details (plan limits, live status, safety; from AgentGid's data)
DeepSeek Harness by DeepSeek. Category: General-purpose AI agents. Type: open-source plugin-based agent harness (desktop, web and CLI). An agent harness from DeepSeek in which every capability is a plugin. Its agent reads and edits workspace files, runs commands, delegates work and keeps a plan, and it runs as a desktop app, a browser UI or the dsh command-line tool. Note: Public preview; the project warns of breaking changes. Open source (MIT) Pricing (checked 2026-11-21): Free and open source (MIT); users enter their own DeepSeek API key or another OpenAI-compatible provider. Free tier: yes. Interfaces: Desktop app, Web app, CLI, Self-hosted Supports: multi_agent Data: Gid Score 67/100, #16 overall; #2 in its category; 284,792.0 GitHub stars; usage +59% over 30 days.

## Alternatives in the same category
- OpenClaw: https://agentgid.com/agents/openclaw/agent.md
- Hermes Agent: https://agentgid.com/agents/hermes-agent/agent.md
- Manus: https://agentgid.com/agents/manus/agent.md
- Cloudflare OS: https://agentgid.com/agents/cloudflare-os/agent.md
- Muse: https://agentgid.com/agents/muse/agent.md
- Andee: https://agentgid.com/agents/andee/agent.md

## Links
- AgentGid page: https://agentgid.com/agents/deepseek-harness/
- Source: https://github.com/deepseek-ai/deepseek-harness
- Source: https://www.deepseek.com/en/harness/
- Source: https://registry.npmjs.org/@deepseek-ai/dsh/latest

---
AgentGid's catalog is curated: every entry was checked against the maker's official pages (registry, repository, license and pricing pages) on the date shown; a license value shows the quote it stands on wherever AgentGid has one. Usage numbers come from AgentGid's own scan of public repositories set up for AI coding agents. Claims by makers are reported, not tested by AgentGid.

## DeepSeek Harness: open-source agent for code and office work

DeepSeek Harness is a free, open-source AI agent from DeepSeek. You run it as a desktop app or in your browser, and it works on files in a folder you choose. It can fix code, run tests, work with documents, spreadsheets and slides, and run tasks on a schedule. You need a DeepSeek account with prepaid credit or your own key from another model provider.

### Where to use it
- Fix bugs and run tests: Point it at your project folder and ask it to find why a test fails; it reads the code, edits files and runs the tests.
- Sort out documents and spreadsheets: Hand it a folder of reports or a spreadsheet; it organizes the files, analyzes the numbers and drafts a document or slides you can preview and refine.
- Weekly report on a schedule: Ask it in plain words to sum up your project notes every Friday afternoon; it sets up a repeating reminder that starts the task on its own.
- Build a small add-on by chatting: In Creator mode you describe a tool you want, such as a focus timer, and the agent writes and installs it as a plugin.
- Run the agent from your Python program: Your own Python script starts the agent on a folder, gives it a task like fixing failing tests and reads back its final answer.

### How to start
- Download the desktop app for a Mac with Apple silicon or for 64-bit Windows from the DeepSeek Harness page.
- Or install Node.js and start the browser version with one command; it opens at http://127.0.0.1:3080.

```
npx @deepseek-ai/dsh web
```

- In Settings → Models enter a DeepSeek key (or add another provider), then choose the project folder as the workspace and start a session.

### License
The code is MIT-licensed, so you may use it at work and in paid products for free; keep the license notice when you ship its code. The model behind it costs money, paid to DeepSeek or to the provider whose key you add. If you publish what it produces, you must mark it as made by AI.

### Good to know
- It is an early preview that changes fast and has had no security audit. DeepSeek warns that updates will break things and that it is not ready for production.
- It runs commands and changes files on your computer. The default setting lets it write only inside the chosen folder and asks you first; a full-access setting skips both.
- DeepSeek's own models need a DeepSeek account and prepaid credit. With another provider's key you need no DeepSeek account, and that provider bills you.
- It works with other providers too, such as Anthropic, OpenAI, Kimi and GLM, or any company gateway that speaks a common protocol. Saved keys stay in a file on your computer and cannot be shown again.

Checked 2026-11-18 against: https://www.deepseek.com/en/harness/, https://www.deepseek.com/en/harness/privacy/, https://www.deepseek.com/en/harness/terms-of-use/, https://github.com/deepseek-ai/deepseek-harness/blob/master/README.md, https://github.com/deepseek-ai/deepseek-harness/blob/master/LICENSE, https://github.com/deepseek-ai/deepseek-harness/blob/master/SAFETY.md, https://deepseek-harness.github.io/deepseek-harness/llms.txt, https://deepseek-harness.github.io/deepseek-harness/en/guide/quickstart.md, https://deepseek-harness.github.io/deepseek-harness/en/guide/providers.md, https://deepseek-harness.github.io/deepseek-harness/en/guide/python-sdk.md, https://deepseek-harness.github.io/deepseek-harness/en/guide/schedule.md, https://deepseek-harness.github.io/deepseek-harness/en/guide/plugin-market.md, https://deepseek-harness.github.io/deepseek-harness/en/guide/github-review.md, https://deepseek-harness.github.io/deepseek-harness/en/guide/public-deployments.md, https://deepseek-harness.github.io/deepseek-harness/en/reference/subsystems/permission-presets.md, https://deepseek-harness.github.io/deepseek-harness/en/reference/subsystems/sandbox.md, https://deepseek-harness.github.io/deepseek-harness/en/reference/subsystems/approval.md, https://registry.npmjs.org/@deepseek-ai/dsh/latest, https://pypi.org/pypi/deepseek-harness-sdk/json
