/mission for Claude Code

Medley: The Multi-Agent Powerhouse for Claude Code

In the evolving landscape of AI-driven development and project management, the ability to execute complex tasks efficiently is paramount. Medley emerges as a game-changing Claude Code plugin designed to bridge the gap between simple prompts and comprehensive project execution. Backed by Y Combinator, Medley is built on the fundamental principle that a coordinated team of agents outperforms a single model working alone. By integrating directly into your terminal, Medley transforms how developers and builders interact with AI through its innovative /mission command.

What's Medley?

Medley is a specialized plugin for Claude Code that introduces a multi-agent orchestration layer to your workflow. Unlike standard AI interactions where a single model attempts to solve a multifaceted problem, Medley decomposes your prompt into a structured, visible plan. It then deploys a coordinated team of agents to tackle these tasks—whether sequentially, in parallel, or even overnight.

At its core, Medley is designed for those who live in the terminal. It doesn't require you to babysit a separate application; instead, it lives inside the Claude Code harness you already use. When you invoke a mission, Medley provides a unique URL that allows you to watch the structure of the work unfold on a dedicated Mission Board in your browser. This transparency ensures that users can track task handoffs, status updates, and the overall progression of complex missions.

Key Features of Medley

Medley is packed with features that prioritize coordination, visibility, and power. Here are the primary features that make the Medley plugin an essential addition to Claude Code:

Multi-Agent Decomposition

The hallmark of Medley is its ability to break down a single prompt into a detailed multi-agent plan. When you type /mission, Medley analyzes the work and creates a task tree. This decomposition allows different agents to focus on specific sub-tasks, ensuring higher accuracy and better results than a single model could achieve alone.

The Mission Board URL

For every mission initiated, Medley generates a URL. This link opens a visual Mission Board in your web browser. This board mirrors the terminal output, showing you the exact task tree, the status of each agent, and how work is being handed off between them. It provides a high-level overview of complex work that would be difficult to track in a standard terminal scroll.

Model Versatility

Medley is model-agnostic within the Claude Code environment. While it is a Claude Code plugin, it allows you to bring a whole team of models to your mission, including:

• Claude
• GPT
• Gemini
• Grok
• Qwen
• DeepSeek
• Ollama
• Llama

Seamless Tool Integration

Medley reaches far beyond code. It integrates with over 40 popular apps and tools, allowing agents to perform research, outreach, documentation, and more. Supported integrations include:

• Development: GitHub, GitLab, VS Code, Vercel, AWS, Supabase, Docker, Netlify, Cloudflare.
• Communication: Slack, Discord, Gmail, Microsoft Teams.
• Productivity: Notion, Linear, Jira, Confluence, Figma, Google Drive, Google Calendar, Obsidian, Raycast.
• Data & Finance: PostgreSQL, MongoDB, Airtable, HubSpot, Stripe.
• Operations: Zapier, Asana, Trello, Sentry.

Built-in Guardrails

Safety and autonomy are balanced in Medley. Agents operate within specific guardrails, and the system is programmed to ask for your permission before performing sensitive actions, such as publishing content, sending outbound communications, spending money, or contacting individuals.

Why Multi-Agent Beats a Single Model

Medley's philosophy is backed by hard data. In various professional benchmarks, the multi-agent approach used by Medley consistently outperforms the highest reported scores from individual models in the Anthropic and OpenAI families.

Benchmark Performance Scores:

• Terminal-Bench 2.1 (Task pass rate): 93.25% (+1.35 pts vs. prior best)
• HealthBench Professional (Mean rubric score): 71.37 (+5.37 pts vs. prior best)
• ViBench (Plan Pass@1): 93.24% (+5.44 pts vs. prior best)
• DrugDiscoveryBench (Exact-answer pass rate): 60.98% (+9.38 pts vs. prior best)

By running a coordinated plan—utilizing parallel agents or different models for specific steps—Medley ensures that the most complex coding and research missions are completed with superior precision.

Use Case Examples

Medley is versatile enough to handle diverse professional missions. Here are several ways you can utilize the /mission command:

• GTM Strategy: Type /mission Find a GTM channel with real signal. Medley will assign agents to research channels, draft outreach, and build lead lists.
• Product Launch: Use Medley to manage a "Waitlist then launch" mission, coordinating docs, social media outreach, and technical deployment.
• Engineering Recruitment: Execute a /mission hire first engineer to scout candidates and organize interview criteria.
• Business Growth: Use missions like "Get my first 100 customers" or "Fix activation" to have a team of agents analyze your funnel and implement changes.

How to Use Medley

Installing and using Medley within Claude Code is a straightforward process that takes only three commands.

Installation Steps

1. Add the Medley marketplace:

   /plugin marketplace add Spine-AI/medley
   

2. Install the plugin:

   /plugin install medley
   

3. Reload your plugins:

   /reload-plugins
   

Starting a Mission

Once installed, you can invoke Medley at any time from your terminal. Simply type /mission followed by your prompt.

Example: /mission get me my first 100 customers

Medley will then begin breaking down the work into a visible plan, which you can monitor via the provided URL or directly in the terminal.

FAQ

Q: What does /mission actually do? A: When you give Medley a prompt via /mission, it creates a full multi-agent plan including targets, budgets, and deadlines. It then runs a coordinated team of agents to execute the steps sequentially or in parallel.

Q: Is Medley a replacement for Claude Code? A: No. Medley is a plugin that makes Claude Code stronger. It adds multi-agent decomposition and parallel execution capabilities to the existing Claude Code environment.

Q: Why does /mission beat a single model? A: It utilizes a coordinated multi-agent plan. This allows for specialized models to handle specific tasks and enables parallel processing, which consistently delivers better results than one model working in isolation.

Q: Does Medley act without my permission? A: No. Medley operates within set guardrails and will always ask for your consent before performing sensitive tasks like sending emails or spending budget.

Q: What does it cost? A: Medley is free to use. It leverages the models you already have access to within Claude Code.