Skip to content

User Manual

Jump to bottom
Harness edited this page May 20, 2026 · 8 revisions

Harness of AI logo

User Manual — Harness of AI

Harness of AI is a meta-agent orchestrator for VS Code: one sidebar for Copilot, Claude, Cursor, Devin, and Kiro, with shared file context and Spec-Driven Development (SDD).

In the app: Ctrl+Shift+P → Harness of AI: Open User Manual (dedicated editor tab).

1. Install

  1. Download harness-vscode-*.vsix from Releases.
  2. Ctrl+Shift+P → Extensions: Install from VSIX...
  3. Developer: Reload Window
  4. Click the Harness of AI icon (fox) in the Activity Bar.

2. Welcome & setup wizard

On first configuration, the welcome screen summarizes what you get in one place:

Feature Description
Unified chat Copilot, Claude, Devin, Cursor, and Kiro in the same panel
Shared context Right-click → Add to Harness of AI Context
Specs (SDD) Skills, Tools, and Workflows in .harness/specs/
MCP External tool servers

Click Get started → to configure agents, or Skip and configure later. The setup wizard ends with the User Manual step before you finish.

3. Chat & shared context

By default Harness attaches: manual context chips, open editor tabs, and the workspace folder (see settings harness.context.includeOpenEditors and harness.context.includeWorkspaceRoot).

  • Explorer or editor → right-click → Add to Harness of AI Context
  • The same context is sent to whichever provider you pick next
  • + New chat / Clear context / Clear Chat & Context

Providers: Auto, Copilot, Claude, Cursor, Devin, Kiro.
Modes: Ask | Agent | Spec+Agent.

Local edits in VS Code (Agent / Spec+Agent)

Mode Copilot Cursor
Ask Chat Cursor Cloud (remote)
Agent Local read/write in workspace Local tools (same as Copilot Agent; needs Copilot token)
Spec+Agent Local + specs Local + specs

Agent mode uses tools: read_file, write_file, list_files, search_in_files, run_git, run_gh. Files change on disk in your project. Cursor Cloud does not edit the IDE tree directly.

For GitHub CLI: configure gh auth login; Agent mode can call run_gh.

4. Configure agents

Open Harness of AI: Open Configuration → Agents tab. Each card shows the provider name, a short description, and Not configured until you add credentials.

Use Configure on each agent, paste the API key or token, then Test Connection. Tokens are stored in VS Code Secret Storage (never in plain-text settings).

Agent Typical use
GitHub Copilot Code review and generation via GitHub
Claude Code Large context and complex reasoning
Devin Autonomous engineering tasks
Cursor AI Cloud Agents API (api.cursor.com)
Kiro (AI-DLC) Kiro CLI + steering rules in .kiro/steering/

5. API servers

In configuration, open the API Servers tab:

  • Built-in agents — default endpoints for Copilot, Devin, and Cursor (read-only rows).
  • Custom API servers — add OpenAI-compatible backends with + Add API server (name, base URL, optional model).

6. Other configuration tabs

Tab Purpose
MCP Connect Model Context Protocol servers (stdio or HTTP)
Workspace Default workspace path, default agent, prompt optimization
Spending Token and request usage per provider

7. Commands

Command Purpose
Harness of AI: Open User Manual This guide (dedicated tab)
Harness of AI: Open Configuration Agents, MCP, workspace, spending
Harness of AI: Initialize Workspace Create .harness/
Harness of AI: Check getGoat Agent diagnostics

8. Help

Repository: github.com/nbsjunior/harness

Todd of AIDLC Wiki

Home

Getting started

Using Todd of AIDLC

Agents

Advanced

Help

Clone this wiki locally