DeepSeek Harness: Building and Extending AI Coding Agents Training Course

Introduction to DeepSeek Harness

• What DeepSeek Harness is and where it fits
• The "everything is a plugin" design philosophy
• Web UI, CLI, and headless modes

Installation and First Run

• Node.js requirements and package installation
• Launching the Web UI with npx
• Building and running from source

Model Configuration

• Setting the DeepSeek API key
• Selecting and switching models
• Adding OpenAI-compatible endpoints

Workspaces and Sessions

• Choosing and switching workspaces
• The session composer and conversation flow
• Reading, editing, and generating files

Running Coding Tasks

• Summarizing repositories and navigating code
• Running commands and delegating work
• Maintaining a task plan

Permissions and the Sandbox

• Approval and permission policies
• Confining file and command access
• Local and remote sandbox backends

Built-in Tools and Integrations

• Filesystem, terminal, and shell tools
• Model Context Protocol support
• Language Server Protocol integration

Subagents and Delegation

• Spawning and forking subagents
• Product providers for other coding agents
• Delegating work across a session

The Cordis Plugin Model

• Plugins as services, events, and effects
• The shared context and lifecycle
• Why there is no privileged core

Profiles, Bundles, and Configuration

• Profiles and bundle layering
• Patching configuration with cordis.patch.yml
• Inspecting the booted tree

Core Subsystems

• The append-only session log
• System prompt and tool schema assembly
• The agent loop and turn flow

Events and Extension Points

• Session, agent, and capability events
• Waterfall and interceptor events
• Mapping behavior to extension points

Developing Plugins

• Registering services on context keys
• Adding tools and model adapters
• Publishing and discovering plugins

SDK, Headless, and Automation

• Driving the harness from another process
• The TypeScript and Python SDKs
• CLI and headless automation