DeepSeek Harness CLI
DeepSeek Harness is an open-source AI agent from DeepSeek where every feature is a plugin. It runs in the browser, on macOS and on Windows.
- 248,935 ★ on the repository
- Last commit
Broad access
Why this level
- Runs code and commands proposed by the model
- Loads third-party plugins with access to files, network and credentials
- No security audit, still a developer preview
Install
Manual install
npx @deepseek-ai/dsh webRequires Node.js. Opens the web UI at 127.0.0.1:3080.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Runs code and commands proposed by the model
- Loads third-party plugins with access to files, network and credentials
- No security audit, still a developer preview
This is third-party code. Review the repository files before installing.
What it does
DeepSeek Harness (the `dsh` command) is an open-source agent released by DeepSeek itself. It uses an everything-is-a-plugin architecture on top of the Cordis framework: you can install ready plugins, write your own, and find community ones under the `dsh-plugin` GitHub topic. The agent starts with `npx @deepseek-ai/dsh web` and opens a web UI at 127.0.0.1:3080. The DeepSeek site offers desktop builds for Apple silicon macOS and 64-bit Windows.
Who it is for. Developers and enthusiasts who want to assemble their own agent from plugins or try DeepSeek's official agent.
Good fit when
- You want to try DeepSeek's official agent
- You want to assemble an agent from plugins for your task
- You need an open-source agent that runs in the browser
Not a fit when
- You need a stable agent for production work: Harness is still a preview
- You cannot let an agent run commands on your main machine
Example request
Run DeepSeek Harness in a separate folder and build a plugin that summarizes the project filesLimitations
The project is a developer preview, and the README warns about breaking changes between versions.
According to SAFETY.md the agent runs model-generated code and commands, loads third-party plugins and can reach the network, processes, files and credentials made available to it. No security audit has been done.
How to disable. Stop the `dsh` process or quit the desktop app. Plugins are removed in the agent UI.
FAQ
Is DeepSeek Harness free?
The agent itself is free under the MIT license. The model is connected separately, and API access is billed by the provider.
How do I install DeepSeek Harness?
Install Node.js and run `npx @deepseek-ai/dsh web`; the web UI opens at 127.0.0.1:3080. The DeepSeek site has desktop builds for Apple silicon macOS and Windows.
Does DeepSeek Harness run on Linux?
Through `npx @deepseek-ai/dsh web` or a source build. The DeepSeek site has no desktop builds for Linux or Intel Macs.
Is DeepSeek Harness safe?
According to SAFETY.md it has had no security audit and can run commands from the model and third-party plugins. The authors advise a separate environment with minimal privileges.
Where do I find plugins for DeepSeek Harness?
Community plugins use the `dsh-plugin` GitHub topic, and you can create your own inside the agent.
Discussion
No questions or reviews yet.
Related
CLI toolsA CLI that keeps an existing agent on a long task: plan a step, run it, verify the result in the real environment and checkpoint progress
A terminal coding agent built on Codex, tuned for low-cost open models like Kimi K3, DeepSeek and Qwen
ByteDance's open agent harness for long tasks: research, code, reports and slides with sub-agents, memory and sandboxes
pydantic-deepagents
pydantic-deepagents (pydantic-deep)
An open-source terminal AI assistant and Python framework with a Docker sandbox, agent tools and subagents
An open-source agent harness in Rust: GUI, CLI, headless and webapp from one binary, with multiple providers, MCP, skills and agent teams
A curated, ranked list of agent harnesses with an MCP server and ready subagents to pick a harness for a task
README in short
The README covers running via npm and from source, the Cordis-based architecture, the developer-preview status and a link to SAFETY.md with safe-use rules.