cami

CAMI Plugin

Claude Capability Management Interface

Your agent scout for Claude Code. Build championship teams of specialized AI agents for any project.

claude plugin install lando-labs/cami

---

What is CAMI?

CAMI is your agent scout and orchestrator. Instead of manually copying agent files between projects, CAMI helps you discover, deploy, and manage teams of specialized Claude Code agents.

Think of it as talent management for AI specialists:

• Create custom agents tailored to your projects
• Browse community sources when you need them
• Get recommendations based on your tech stack
• Deploy agents to projects with one command
• Keep your roster updated across all projects

Skills-only architecture: No binary installation, no MCP server configuration. Just install the plugin and start talking.

Zero-config: Install and go. CAMI creates its workspace automatically when you create your first agent, add your first source, or deploy your first agent.

---

Glossary

New to CAMI? Here's what the terms mean:

Term	What It Means
Agent	A markdown file that gives Claude specialized expertise. Think of it as a persona or specialist you can call on - like a "React performance expert" or "API design specialist".
Skill	A workflow automation for Claude. While agents provide expertise, skills provide step-by-step procedures - like "run tests before commit" or "create PR with changelog".
Source	A folder or Git repository containing agents and skills. Like a talent pool you can recruit from.
Workspace	CAMI's home base (~/.claude/cami-workspace/ by default). Contains your sources, config, and custom agents.
Deploy	Copy an agent or skill to a project's .claude/ folder so it's available in that project.
Roster	The agents currently deployed to a specific project. Your project's "team lineup".
Scout	CAMI's personality - it analyzes your project and recommends agents that fit your tech stack.

---

Quick Start (Zero-Config)

Minute 1: Install and say hi

claude plugin install lando-labs/cami

Then in any Claude Code session:

You: "Hey CAMI"

CAMI: "Hey! I'm CAMI - your agent scout. I help you build championship teams
      of AI agents for your projects.

      I notice you haven't set up your workspace yet. This is your command center
      for managing agent sources - think of it as your talent network.

      Ready to set it up? [y/n]"

Minute 2: What are you building?

Option A: Create a custom agent (Recommended)

You: "I'm building a React dashboard with complex state management"

CAMI: "Perfect! Let me create a custom agent for you.

      I'll make a state-management-specialist that understands:
      - React Context vs Redux vs Zustand trade-offs
      - Performance optimization patterns
      - Testing stateful components

      Sound good?"

You: "Yes"

CAMI: "Creating your workspace at ~/.claude/cami-workspace/
      Creating state-management-specialist in sources/my-agents/... Done!

      Ready to deploy it to this project?"

Option B: Browse community sources

You: "Yes. Scout this project."

CAMI: "Done! Created your workspace at ~/.claude/cami-workspace/

      Looking at your React/TypeScript project... found package.json with React 19.

      For this stack, I'd recommend creating:
      - A frontend-methodology agent (guides UI architecture decisions)
      - A react-patterns skill (React best practices)

      Want me to create these for you? Or browse community sources first?"

Minute 3: Agents deployed, you're working

You: "Yes, create them"

CAMI: "Creating frontend-methodology in my-agents...
      Creating react-patterns skill...
      Deploying to my-app... Done!

      Your roster is set - ready to build.
      Try @frontend-methodology to start planning your architecture."

No config files. No setup commands. Just conversation.

---

Installation

From Marketplace (Recommended)

claude plugin install lando-labs/cami

Local Development

git clone https://github.com/lando-labs/cami-plugin.git
claude --plugin-dir /path/to/cami-plugin

---

Available Skills

CAMI operates through conversational skills. You don't need to invoke them explicitly - just talk naturally.