Deployment Options for Ponytail: Install in 6 AI Agent Environments

Ponytail deploys as a Node.js plugin across six major AI-agent platforms—Claude Code, Codex, GitHub Copilot CLI, Pi agent harness, OpenCode, and Gemini CLI—using marketplace commands or JSON configuration.

The DietrichGebert/ponytail repository distributes Ponytail as a Node.js plugin designed to integrate with various AI-agent environments. Understanding the available deployment options for Ponytail ensures you can activate its lifecycle hooks and skills within your preferred coding assistant. Each target platform uses specific installation commands to register the plugin from the repository's hooks/ and skills/ directories.

Supported Deployment Targets

Ponytail supports six distinct AI-agent environments. Select your platform below for the exact installation sequence.

Claude Code

Claude Code integrates Ponytail through its plugin marketplace. The installation registers two lifecycle hooks that expose Ponytail's commands under the /ponytail namespace.

/plugin marketplace add DietrichGebert/ponytail
/plugin install ponytail@ponytail

Once installed, commands appear as /ponytail:<command> within the Claude Code interface.

OpenAI Codex

For Codex desktop clients, the plugin loads its hooks from the repository's hooks/ directory after marketplace registration.

codex plugin marketplace add DietrichGebert/ponytail
codex plugin add ponytail@ponytail

GitHub Copilot CLI

Copilot CLI users install Ponytail via the plugin marketplace, enabling slash shortcuts in interactive sessions.

copilot plugin marketplace add DietrichGebert/ponytail
copilot plugin install ponytail@ponytail

After installation, invoke skills using /ponytail:<command> during Copilot sessions.

Pi Agent Harness

The Pi harness clones the repository directly and registers Ponytail as a built-in agent.

pi install git:github.com/DietrichGebert/ponytail

This method makes Ponytail available through the /ponytail invocation path without separate marketplace steps.

OpenCode

OpenCode requires JSON configuration in your opencode.json file to load the plugin from npm or local checkouts.

{
  "plugin": ["@dietrichgebert/ponytail"]
}

OpenCode automatically injects the Ponytail ruleset and commands upon loading the configuration.

Gemini CLI

Gemini CLI mirrors the Codex workflow, enabling Gemini-based agents to access Ponytail's capabilities.

gemini plugin add github.com/DietrichGebert/ponytail

Consult the repository README for the latest syntax variations, as flags may differ across Gemini CLI versions.

Core Plugin Architecture

All deployment options share identical core assets located in the repository root. The hooks/ directory contains Node.js lifecycle scripts that execute before and after each AI turn, while the skills/ directory houses individual "SKILL" files such as ponytail-review and ponytail-debt that implement concrete capabilities.

The plugin.yaml and plugin.json files define metadata used by Copilot, Codex, and Claude agents to identify and load the plugin correctly.

Prerequisites and Environment Setup

Node.js must be present on the system PATH for Ponytail's lifecycle hooks to execute. As documented in the repository's README.md Install section, missing Node.js installations result in silent failures: the skills remain technically usable, but the always-on activation hooks will not trigger.

Verify Node.js availability before running any installation commands:

node --version

Summary

• Ponytail deploys to six AI-agent platforms: Claude Code, Codex, GitHub Copilot CLI, Pi harness, OpenCode, and Gemini CLI.
• Claude Code, Codex, and Copilot CLI use marketplace addition commands followed by plugin installation.
• Pi harness clones directly from GitHub using the pi install command.
• OpenCode requires JSON configuration in opencode.json to load @dietrichgebert/ponytail.
• All deployments rely on the hooks/ and skills/ directories and require Node.js on the PATH for full functionality.

Frequently Asked Questions

Does Ponytail support self-hosted or private AI agents?

Ponytail is designed for the six documented platforms (Claude Code, Codex, Copilot CLI, Pi, OpenCode, and Gemini CLI). The architecture depends on specific lifecycle hooks in the hooks/ directory and metadata in plugin.yaml/plugin.json, which align with these supported environments. Custom agent integrations would require adapting these hook structures.

What happens if Node.js is not installed when deploying Ponytail?

If Node.js is missing from the PATH, the lifecycle hooks in the hooks/ directory cannot execute. According to the README.md documentation, the plugin's skills remain accessible, but the always-on activation features remain silent, preventing automatic triggering before and after AI turns.

Can I use Ponytail in multiple AI agents simultaneously?

Yes. Since Ponytail is distributed as a Node.js plugin via the DietrichGebert/ponytail repository, you can install it independently in Claude Code, Codex, and other supported environments on the same machine. Each installation references the same skills/ and hooks/ assets but operates within its respective agent's isolated context.

Where are the installation commands documented in the repository?

The primary installation steps for each deployment target reside in README.md. Specific line ranges include the Claude Code instructions at lines 30-38 and the OpenCode configuration section at lines 80-88.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}