cloudbees-unify

CloudBees Unify Agent Plugin for VS Code, Claude Code, and Copilot CLI

This repository provides a prepackaged agent plugin that lets AI coding agents (VS Code Copilot Chat, Claude Code, and Copilot CLI) talk to CloudBees Unify — including components, workflow runs, build status and logs, security findings, feature flags, CI controllers, users, and teams — via the Unify MCP server.

Status: Scaffold. The example agent and skill are placeholders meant to be replaced as real capabilities are added.

Installation

VS Code (Agent Plugins)

Recommended: Install as an Agent Plugin in VS Code 1.99+ (Copilot Chat agent mode):

1. Open the Extensions view (⇧⌘X) and search for @agentPlugins.

2. If published, search for cloudbees-unify and click Install.

3. To install from source, run Chat: Install Plugin From Source from the Command Palette and enter:

   https://github.com/cloudbees/unify-ai-plugins.git
   

4. The plugin will appear in the Agent Plugins - Installed section. Enable it if needed.

Manual (local clone):

1. Clone this repo:

   git clone https://github.com/cloudbees/unify-ai-plugins.git
   

2. Register the plugin in your VS Code settings.json:

   "chat.pluginLocations": {
     "/absolute/path/to/unify-ai-plugins": true
   }
   

3. Reload VS Code. The plugin will be enabled and its skills, agents, hooks, and MCP servers will be auto-discovered.

Claude Code

Install from the Claude Code marketplace (once published):

/plugin marketplace add cloudbees/unify-ai-plugins
/plugin install cloudbees-unify@unify-ai-plugins

Or install locally from a clone:

git clone https://github.com/cloudbees/unify-ai-plugins.git
claude plugin install ./unify-ai-plugins --scope user

GitHub Copilot CLI

Uses the same plugin format as above. See GitHub Copilot CLI plugin reference.

How it works

This plugin ships a .claude-plugin/plugin.json (Claude/VS Code compatible) and a .mcp.json that points at the CloudBees Unify MCP server (https://mcp.cloudbees.io/v1/mcp).

When enabled, VS Code, Claude Code, and Copilot CLI auto-discover:

• Skills (in skills/)
• Agents (in agents/)
• Hooks (in hooks/ or hooks.json)
• MCP servers (from .mcp.json)

On first use, the MCP server opens a browser window for OAuth (Google, GitHub, or SSO). After that, agent tools invoke Unify transparently.

Usage

After installing and enabling the plugin, just ask in plain English. Claude Code, Copilot Chat, or Copilot CLI will route your request through the unify-assistant sub-agent and the appropriate skill.

Example requests

What components are in my CloudBees Unify org?
Show me the latest build status for the checkout-service component.
Are there any open critical security findings this week?
List feature flags for the storefront application.
Why did the last workflow run on payments-api fail? Show me the logs.

The first command in a session opens a browser for OAuth (Google / GitHub / SSO). After that, calls run silently.

Invoke a sub-agent explicitly

Use the @ prefix to scope a request to the Unify assistant:

@unify-assistant list components in the platform-eng organization

Invoke a skill directly

Skills are addressable as /cloudbees-unify:list-components:

/cloudbees-unify:list-components
/cloudbees-unify:list-components organization=platform-eng filter=api

Chain Unify into a coding task

The point of running this in Claude Code or Copilot Chat is mixing Unify data with code:

Find the component that owns this repo in Unify, then show me the last 5 builds
  and open a diff between the last green commit and HEAD.

A Snyk finding flagged lodash in the auth-service. Pull the finding from Unify,
  then update package.json and the lockfile to the patched version.

Layout

• .claude-plugin/ — plugin.json + marketplace.json
• .mcp.json — Unify MCP server config (auto-wired on install)
• agents/ — sub-agent definitions
• skills/ — skill definitions invoked by sub-agents or directly
• hooks/ — SessionStart hook + system prompt injected on session start

See CONTRIBUTING.md for how to add agents and skills.

Troubleshooting

• If the plugin does not appear after installation, check that the name field in plugin.json uses only lowercase letters, numbers, and hyphens.
• If skills do not load, ensure the skill directory and SKILL.md frontmatter use plain kebab-case names.
• If you see an OAuth prompt, this is expected on first use.
• For more, see the Agent Plugins for VS Code documentation.

License

MIT — see LICENSE.