Getting Started with cAgents

Most AI tools give you one assistant for one task. You describe what you want, it produces something, you edit it yourself. That model works fine for simple requests. It starts breaking down when tasks are genuinely complex — when "build me a feature" really means plan, design, implement, test, and document, where each phase depends on the last.

cAgents is a multi-agent orchestration framework for Claude Code that handles this kind of work. Instead of one AI doing everything, cAgents coordinates specialized agents — planners, executors, reviewers, domain leads — each focused on what it does best. You describe the outcome you want. cAgents figures out which agents to involve, in what order, and coordinates the work.

By the end of this article you've installed the plugin, authored a CLAUDE.md for your project, and picked your first task. The prompts panel on the right walks each step end-to-end with branching for the cases the prose doesn't cover.

Install the plugin

cAgents runs on top of Claude Code, Anthropic's CLI for running AI agents in your terminal — install that first (see Claude Code Quickstart in the official docs). cAgents itself ships as a Claude Code plugin (see Discover Plugins in the official docs).

Then install cAgents:

/plugin install CaelanDrayer/cAgents

That's it. Claude Code pulls the plugin, registers the skills, and makes them available immediately. No separate activation step.

Verify the install:

/cagents:helper "What commands are available?"

If the install worked, cAgents responds with a summary of available commands. The full set — /designer, /run, /team, /org, /optimize, /debug, /review — is now available in any project directory.

Author your CLAUDE.md

Before running any cAgents commands, create a CLAUDE.md at your project root (see Memory in the official docs for the canonical reference, or claude-code-guide Part 1 for the in-series treatment). The five-anchor template below adds a cAgents-specific lens — what makes a CLAUDE.md work well for cAgents commands — on top of those references. This file is how you give agents persistent context about your project — what it is, how it's structured, what conventions to follow.

Without a CLAUDE.md, agents have to infer everything from the files they can read. With one, they start with a clear picture of what they're working in.

What goes in CLAUDE.md

A good CLAUDE.md covers five anchors:

• Project purpose — one or two sentences on what this project is.
• Structure — where things live and why.
• Conventions — naming, file organization, key patterns to follow.
• Technology — the main stack and any constraints.
• What not to touch — files or directories that should be off-limits.

Two shapes for two project types:

# CLAUDE.md

## Project
E-commerce site for handmade goods. Built with Next.js 14, TypeScript, and Stripe for payments.

## Structure
- /app - Next.js App Router pages
- /components - shared React components
- /lib - utilities and Stripe integration
- /content - product data as markdown files

## Conventions
- Components: PascalCase, one per file
- Utility functions: camelCase, exported individually
- Never modify /lib/stripe.ts without reviewing the test suite first

## Stack
Next.js 14, TypeScript (strict), Tailwind CSS, Stripe, Vercel deployment

# CLAUDE.md

## Project
Content marketing strategy for a B2B SaaS company selling HR software to mid-market companies.

## Target Audience
HR directors and VPs at companies with 200-2000 employees. They care about compliance,
reducing manual work, and making the case to CFOs.

## Content Pillars
1. Compliance and regulation
2. HR automation and efficiency
3. Employee experience

## Voice
Professional but direct. No buzzwords. Cite sources. Avoid "leverage" and "synergy."

## Deliverables
Blog posts, email sequences, LinkedIn content, and one-pagers for sales.

The prompts panel turns this into a structured interview that walks each anchor, picks the technical-or-non-technical template based on what's actually in your project root, and writes the file when you approve.

When to reach for which command

What's next

This series covers each command in depth, with step-by-step walkthroughs following the same two projects through every stage:

1. Getting Started ← you are here
2. /designer — Design Before You Build — Interactive Q&A that turns vague requirements into implementation-ready specs. Start here before any non-trivial task.
3. /run — Execute Tasks with Agents — The workhorse. Give it a task, it plans and executes. Covers the full pipeline, key flags, and follow-up patterns.
4. /team — Parallel Multi-Agent Execution — Wave-based parallel execution for multiple independent tasks. When and why to use it over /run.
5. /org — Cross-Domain Orchestration — C-suite agents coordinating across business domains. For work that genuinely spans engineering, marketing, legal, or other domains.
6. /optimize — Performance and Efficiency — Measurable improvements with before/after metrics and rollback safety.
7. /debug — Systematic Debugging — 4-phase debugging for issues that defy quick fixes.
8. /review — Quality Review — Parallel specialist review before you ship.
9. Sessions — Under the Hood — How cAgents tracks every pipeline run on disk.
10. Hooks — The Event System — The live event-driven machinery behind every pipeline.

If your project has clear requirements and you're ready to go: start with Part 3: /run. If you want to think through the design first: start with Part 2: /designer.