MuleRouter Agent Skill

Agent Skill for calling MuleRouter / MuleRun multimodal APIs to generate images, videos, speech, and music. The skill is a thin documentation layer over the mulerouter npm CLI, which handles all network calls, image base64 conversion, and async task polling.

Features

• Multiple sites — supports both MuleRouter and MuleRun gateways
• 37+ endpoints — Wan2.1/2.2/2.5/2.6, Happy Horse, Seedance 2.0, Nano Banana / Pro, Veo3, GPT-Image-2, Kling V3 / V3 Omni, Midjourney, MiniMax Speech 2.8 + Music 2.5
• Single CLI surface — mulerouter run // --flag value …
• Easy configuration — environment variables or .env file
• Async task handling — automatic polling, or split into --no-wait + mulerouter status for long jobs
• AI-friendly discovery — mulerouter list / mulerouter params enumerate everything

Installation

Step 1 — Install the CLI

npm install -g mulerouter
# or run ad-hoc without installing
npx -y mulerouter@latest --help

Requires Node.js 18 or later.

Step 2 — Install the skill

Via Claude CLI

claude plugin marketplace add openmule/mulerouter-skills
claude plugin install mulerouter-skills

Via Claude Code session

/plugin marketplace add openmule/mulerouter-skills
/plugin install mulerouter-skills

After installation, restart Claude Code to load the new skill.

Configuration

Required Environment Variables

API endpoint (one required)

If both are set, MULEROUTER_BASE_URL wins.

Setup options

Option A — shell env

export MULEROUTER_API_KEY="your-api-key"
export MULEROUTER_BASE_URL="https://api.mulerouter.ai"
# or instead: export MULEROUTER_SITE=mulerouter

Option B — .env file

MULEROUTER_API_KEY=your-api-key

# pick ONE
MULEROUTER_BASE_URL=https://api.mulerouter.ai
# MULEROUTER_SITE=mulerouter   # or: mulerun

The CLI only loads variables prefixed with MULEROUTER_ from .env. Other variables are ignored.

Verify

mulerouter --version
mulerouter config              # show current configuration / diagnose env

Quick Start

Once installed and configured, just ask Claude to use MuleRouter to generate something — the skill self-documents available models. Manually you can also call the CLI directly:

# enumerate endpoints
mulerouter list --tag SOTA

# inspect parameters for one endpoint