Skip to content
| Marketplace
Sign in
Visual Studio Code>Programming Languages>Harness ManagerNew to Visual Studio Code? Get it now.
Harness Manager

Harness Manager

AdmiralGallade

|
37 installs
| (1) | Free
Install and manage AI agent harnesses for VS Code, Cursor, and Windsurf. Instantly configures Claude Code, GitHub Copilot, Cursor, and Windsurf with the right instructions for your workflow.
Installation
Launch VS Code Quick Open (Ctrl+P), paste the following command, and press enter.
Copied to clipboard
More Info

Harness Manager

VS Code Marketplace Open VSX

A lightweight VS Code, Cursor and Windsurf extension for installing and managing AI agent harnesses.

Why use this

  • Install harnesses from a GitHub repository
  • Keep multiple harnesses available with optional multi-install support
  • Automatically update AI agent config files for 10 tools — including the cross-tool AGENTS.md standard, Claude Code, Copilot, Cursor, Windsurf, Cline, Roo Code, Gemini CLI, Aider, and JetBrains Junie
  • Choose exactly which tools to write files for
  • Sync or reset your AI instruction files at any time
  • See the active harness in the status bar
  • Restore previous harness states from automatic backups

Install

  1. Open a folder in VS Code
  2. Open the Harness Manager sidebar
  3. Select a harness and click Install

Add your own harnesses

You can publish your own harness repository or upload harnesses here:

https://github.com/AdmiralGallade/harness-repository

Create a new harness scaffold

Use the Harness Manager: Create Harness Scaffold command to generate:

  • harnesses/<id>/config.json
  • harnesses/<id>/template.yaml
  • harnesses/<id>/README.md
  • harnesses.json with a register entry for the scaffold

This helps you publish a new harness repository quickly.

Configuration

  • harnessManager.githubRepo — GitHub repo to load harnesses from (owner/repo)
  • harnessManager.pointerTargets — which AI tools to generate instruction files for (leave empty for all)
  • harnessManager.defaultCreateLocation — workspace-root or prompt for scaffold and install locations
  • harnessManager.multiHarnessInstall — install without removing existing harnesses
  • harnessManager.cacheRefreshInterval — cache lifetime in milliseconds

What is a harness?

A harness is a collection of files (prompts, rules, hooks, skills) that tells your AI agent how to behave in a project. Installing one writes configuration files that are automatically picked up by:

Tool File(s) written
AGENTS.md (cross-tool standard) AGENTS.md
Claude Code .claude/CLAUDE.md · .claude/active-harness.md
GitHub Copilot .github/copilot-instructions.md
Cursor .cursorrules · .cursor/rules/harness.mdc
Windsurf .windsurfrules · .windsurf/rules/harness.md
Cline .clinerules
Roo Code .roo/rules/harness.md
Gemini CLI GEMINI.md
Aider CONVENTIONS.md
JetBrains Junie .junie/guidelines.md

AGENTS.md is the open, cross-tool standard read natively by most modern coding agents. Use the harnessManager.pointerTargets setting to pick exactly which tools receive files — leave it empty to write for all of them.

Compatible editors

  • VS Code — install from the VS Code Marketplace
  • Cursor, Windsurf, VSCodium & other VS Code forks — install from Open VSX, or side-load the .vsix

Also available on the Open VSX Registry.


Features

  • Browse harnesses from a GitHub repository, grouped by category
  • Install with one click — files are copied into agent-harnesses/ with full directory structure preserved
  • Star harnesses to pin them above the list for quick access
  • Focus mode — collapse everything except your starred harnesses
  • Active section — all installed harnesses are pinned at the top
  • Multi-harness support — optionally keep several harnesses installed side by side
  • Import a harness from a local folder or ZIP file
  • Multi-tool output — write instruction files for up to 10 tools, including the AGENTS.md cross-tool standard; pick which ones via settings
  • Sync — regenerate all AI instruction files from the active harness (fixes drift or applies a changed tool selection)
  • Reset — clear all AI instruction files back to a placeholder and deactivate the current harness
  • Status bar — the active harness is shown in the status bar; click it to open the sidebar
  • Version history — every install or switch creates an automatic backup; restore any previous state with one click
  • Remove any installed harness (backup is saved first)
  • Search the full list, or only your starred harnesses in focus mode
  • Use your own repository — point the extension at any public GitHub repo that follows the harness format

Getting Started

  1. Install the extension from the VS Code Marketplace
  2. Open a folder in VS Code
  3. Click the Harness Manager icon in the activity bar
  4. Click Install on any harness
  5. The harness is copied into agent-harnesses/ and your AI config files are updated immediately

Sidebar Layout

✓ Active Harnesses     ← all currently installed harnesses
★ Starred              ← your favourites (not yet installed)
  Available Harnesses  ← everything else, grouped by category
  Import Harness       ← import from a local folder or ZIP
  Harness Repository   ← switch to a different GitHub repo
  Version History      ← backups with restore and clear-all

Settings

Setting Default Description
harnessManager.githubRepo AdmiralGallade/harness-repository GitHub repo to load harnesses from (owner/repo)
harnessManager.pointerTargets (all 10 tools) Which AI tools to generate instruction files for; leave empty for all
harnessManager.multiHarnessInstall false When on, installing a harness does not remove others
harnessManager.activeHarnessId (auto) ID of the primary active harness (set automatically)
harnessManager.cacheRefreshInterval 86400000 Cache lifetime in ms (default 24 h)

Commands

Command Description
Harness Manager: Sync AI Instruction Files from Active Harness Regenerate instruction files for the active harness
Harness Manager: Reset AI Instruction Files (Clear Active Harness) Reset instruction files to a placeholder and clear the active harness
Harness Manager: Switch Harness Repository Point the extension at a different GitHub repo
Harness Manager: Create Harness Scaffold Generate a starter harness for publishing
Harness Manager: Refresh Harness List Re-fetch the harness catalog

Harness Repository Format

Point the extension at any public GitHub repo with this structure:

harness-repository/
├── harnesses.json          # manifest listing all harnesses
└── harnesses/
    └── my-harness/
        ├── config.json
        ├── template.yaml
        └── README.md

harnesses.json schema:

{
  "version": "1.0",
  "lastUpdated": "2026-01-01T00:00:00Z",
  "harnesses": [
    {
      "id": "my-harness",
      "name": "My Harness",
      "description": "What this harness does",
      "category": "Development",
      "tags": ["typescript", "testing"],
      "dependencies": [],
      "author": "Your Name",
      "version": "1.0.0",
      "files": [
        {
          "path": "harnesses/my-harness/template.yaml",
          "type": "template",
          "description": "Main instructions"
        }
      ]
    }
  ]
}

License

MIT

  • Contact us
  • Jobs
  • Privacy
  • Manage cookies
  • Terms of use
  • Trademarks
© 2026 Microsoft