Skip to content
HUIAHUB

deepseek-ai/deepseek-harness

deepseek-harness

DeepSeek Harness: Everything is a Plugin.

Metrics updated Emerging+2.3% 7d
View on GitHub (opens in a new tab)Homepage / demo

GitHub metrics

Stars
249K
+5.7K this week
Forks
30K
30-day growth
—
Open issues
0
243K246K249KOct 4, 2026Oct 13, 2026
GitHub stars from the official GitHub API, recorded each time the repository was checked (3 days).
GitHub stars by day
DayGitHub stars
2026-10-04242885
2026-10-10246717
2026-10-13248562

What it does

DeepSeek Harness is an open-source agent runtime from DeepSeek AI that runs coding agents through a Web UI, headless jobs, SDK connections or ACP automation clients. Its Cordis-based design treats the model adapter, tools, session log and agent loop as plugins, so these parts can be replaced or extended through configuration. The Web UI works with a selected project directory. An agent can read and edit files, execute commands, delegate work and keep a plan, subject to the configured approval policy.

The default setup supports DeepSeek models, while the model settings also support shipped third-party providers and custom APIs using OpenAI Chat Completions, OpenAI Responses or Anthropic Messages protocols. Developers can add their own Harness plugins, including plugins that register tools or services. The project is currently an experimental developer preview. Its maintainers warn that compatibility-breaking changes will occur and that it has not received a security audit. Because agents can execute generated commands and access permitted files, processes, credentials and networks, the safety documentation recommends restricted privileges and isolated environments.

Why it matters

DeepSeek Harness gives agent developers one runtime whose major parts are plugins rather than fixed internal components. Models, tools, persistence, the agent loop and other services can be changed through the same composition system. It also provides several execution profiles, including a browser interface, one-shot headless runs, JSON-RPC SDK access and ACP automation, which makes the same agent infrastructure usable from different clients.

Main features

  • Runs an agent through a local Web UI.
  • Reads and edits workspace files and runs commands.
  • Supports Web, headless, SDK and ACP profiles.
  • Uses plugins for models, tools, sessions and the agent loop.
  • Accepts DeepSeek, shipped third-party and custom model providers.
  • Supports OpenAI Chat Completions, OpenAI Responses and Anthropic Messages APIs.
  • Lets TypeScript plugins register tools, services and events.

Who it's for

  • Developers building or modifying coding agents and agent tools.
  • Teams that need custom model providers or OpenAI-compatible model gateways.
  • Plugin authors who want to add tools or services to the Harness runtime.
  • Developers experimenting with Web, headless, SDK or ACP agent workflows.

Strengths

  • Major agent components are replaceable plugins rather than a fixed core.
  • One installation provides Web, headless, SDK and ACP execution profiles.
  • Custom model APIs can use three documented provider protocols.
  • The project includes documentation for writing TypeScript plugins and tools.

Limitations

  • It is experimental developer-preview software with expected compatibility-breaking changes.
  • The project has not undergone a security audit and is not production-ready.
  • Agent commands and plugins can affect accessible files, credentials, processes and networks.
  • Sandboxing and approval controls do not guarantee isolation.

Getting started

For a packaged local Web UI, install Node.js and run:

npx @deepseek-ai/dsh web

To run a repository checkout after cloning it, the README gives these commands:

pnpm install
pnpm run build
pnpm dsh web

Requirements: Node.js is required for the packaged Web UI; model use requires a configured provider credential such as a DeepSeek API key, while source builds use pnpm.

Always follow the official instructions in the README on GitHub. This overview is our summary, not a copy.

Sources