deepseek-harness/docs/user/guide/index.md

Introduction

English | 中文

DeepSeek Harness is a plugin-based agent development framework built on the Cordis microkernel. Its central idea is simple: everything is a plugin.

What it is

Harness implements every capability an AI agent needs—including LLM calls, tool execution, session management, and subtask delegation—as a composable plugin. A cordis.yml file declares which plugins to load and how to configure them, assembling a complete agent.

# Select the LLM backend
- name: '@deepseek-ai/dsh-llm-deepseek'
  config:
    apiKey: !!js process.env.DEEPSEEK_API_KEY

# Select the agent the interactive front door drives
- id: agent-loop
  name: '@deepseek-ai/dsh-agent-loop'
  config:
    agents:
      - id: main
        provider: deepseek-official
        model: deepseek-v4-flash

# Select the interactive front door
- name: '@deepseek-ai/dsh-tui'

Who it is for

Application users

To run an existing agent application, such as a coding assistant or conversational agent:

1. Copy an example template.
2. Add an API key.
3. Run it.

No code is required. See the quick start.

Plugin developers

To add a custom tool, a new LLM adapter, or another execution backend, write a plugin. Harness provides explicit extension interfaces and a type-safe development experience. See development.

Core features

• Configuration only — cordis.yml selects the capability set; changing a model or adding a tool is a configuration edit.
• Hot replacement (HMR) — edit plugin code during development without restarting the process.

Technology

• Runtime: Node.js ^22.19 or >= 24
• Language: TypeScript (ESM)
• Framework: Cordis
• Package manager: pnpm workspaces (the repository pins pnpm 11)