Home
Harness of AI is a meta-agent orchestrator for VS Code: one IDE, many AI providers (Copilot, Claude, Cursor, Devin, Kiro), plus Spec-Driven Development and context engineering in the same sidebar — without switching editors per vendor.
User Manual · Why Harness of AI?
| I want to… | Start here |
|---|---|
| User manual | User Manual |
| Understand why to use Harness of AI | Why Harness |
| Install and send my first message | Getting Started |
| Use the chat UI (providers, modes) | Chat Interface |
| Configure API keys | Configuration |
| Set up GitHub Copilot | Copilot Modes |
| Set up Cursor AI | Agent Connectors → Cursor |
| Use Auto provider | Auto Routing |
| Use Specs (SDD) | SDD Specs |
| Understand prompt optimization | Prompt Optimization |
| Track tokens and requests | User Manual → Spending |
| Run without VS Code (CLI only) | Dual Mode |
| Fix errors | Troubleshooting |
VS Code Extension (UI) ←→ CLI daemon (IPC) ←→ Agent connectors
│
Copilot · Devin · Cursor · Claude · Kiro
- One IDE, many providers — use Copilot, Claude, Cursor, Devin, and Kiro from VS Code; no separate IDE per AI vendor.
-
Spec-Driven Development — versioned specs in
.harness/specs/; Spec+Agent injects them into the prompt. - Context engineering — attach files once; the same context follows you when you switch provider.
- Prompt optimization — built-in pipeline for token efficiency and answer quality on every provider (details).
- Extension — chat sidebar, spec manager, configuration panel
-
CLI — all file I/O, authentication, agent routing (bundled inside the
.vsix) -
Specs — YAML skills/tools/workflows in
.harness/specs/
Repository: github.com/nbsjunior/harness
- Agent Connectors
- Copilot Modes (Ask / Agent / Spec+Agent)
- Architecture
- IPC Protocol
- Dual Mode (Extension vs CLI)
- AI-DLC and Kiro
- Development Guide
If you are an AI model working on this codebase, read AGENTS.md in the repository first.