Open-source desktop app for running a local multi-agent AI workforce that automates multi-step computer tasks, positioned as a free alternative to Claude Cowork.
deepseek-ai/deepseek-harness
An open-source agent harness from DeepSeek AI, built on an everything-is-a-plugin architecture on top of the Cordis framework.
- Stars
- 253,976
- Forks
- 30,706
- Open issues
- 0
- Last commit
- 3 days ago
What is it?
DeepSeek Harness (dsh) is an open-source agent harness developed by DeepSeek AI. Its core design is that every capability is a plugin, and it is built on the Cordis framework, which provides the plugin and composition model.
The project can be started quickly with npx @deepseek-ai/dsh web, which launches a local Web UI at port 3080 by default. Running from a source checkout requires pnpm install and a build step before starting the Web UI. Developers can also publish their own plugins and tag the repository with the dsh-plugin topic for discovery.
The project is in developer preview and changes quickly, so compatibility-breaking changes should be expected. A safety notice is provided and should be reviewed before running it.
Key features
- 01Everything-is-a-plugin architecture built on Cordis.
- 02Launches a local Web UI with a single npx command.
- 03Runnable from npm or from a source checkout with pnpm.
- 04Provides development guides, architecture docs, and an AGENTS.md for coding agents.
- 05Includes desktop build targets via Make alongside the Web workflow.
Use cases
- Running a local agent Web UI for experimentation.
- Building and publishing custom plugins for an agent harness.
- Studying a plugin-first agent architecture built on Cordis.
Quick start
- 1
Install Node.js, then start the Web UI with npx
npx @deepseek-ai/dsh web - 2
Or clone the repository and install dependencies with pnpm
git clone https://github.com/deepseek-ai/deepseek-harness.git && cd deepseek-harness && pnpm install - 3
Build the repository artifacts
pnpm run build - 4
Start the Web UI from the source checkout
pnpm dsh web
Things to know
- The project is in developer preview and will have compatibility-breaking changes.
- Review the safety notice in SAFETY.md before running it.
Topics
FAQ
- What is deepseek-harness?
- An open-source agent harness from DeepSeek AI, built on an everything-is-a-plugin architecture on top of the Cordis framework. DeepSeek Harness (dsh) is an open-source agent harness developed by DeepSeek AI. Its core design is that every capability is a plugin, and it is built on the Cordis framework, which provides the plugin and composition model.
- Is deepseek-harness free to use? What license does it have?
- deepseek-harness is open source under the MIT license. Check the license terms before using it in commercial products.
- Is deepseek-harness still maintained?
- Actively maintained. The last commit to deepseek-harness was on Oct 17, 2026.
- What can I use deepseek-harness for?
- Running a local agent Web UI for experimentation. Building and publishing custom plugins for an agent harness. Studying a plugin-first agent architecture built on Cordis.
- How do I install deepseek-harness?
- Install Node.js, then start the Web UI with npx: npx @deepseek-ai/dsh web Or clone the repository and install dependencies with pnpm: git clone https://github.com/deepseek-ai/deepseek-harness.git && cd deepseek-harness && pnpm install Build the repository artifacts: pnpm run build Start the Web UI from the source checkout: pnpm dsh web
- Where can I get deepseek-harness?
- The source code and installation instructions are on GitHub: https://github.com/deepseek-ai/deepseek-harness
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
Open-source SDK for building and running AI agents in Python and TypeScript, with tools, MCP, memory, and model portability.
A DSH plugin that renders interactive UI components (charts, tables, forms, quizzes, plots, Mermaid) inline inside assistant replies.
Open-source workflow engine that runs YAML-defined AI coding and other processes with agents, scripts and human approval gates.
The official TypeScript SDK for building Model Context Protocol servers and clients that give LLM applications standardized context.
TypeScript framework for building AI agents with memory, tools, MCP, RAG, voice and workflows, plus the VoltOps console for observability.