Meet Your 54 New Teammates: The Complete Claude Code Unified Agents Guide

For developers, DevOps, data scientists, product managers, and anyone who wants expert help on demand.

Table of Contents

  1. What Exactly Is Claude Code Unified Agents?
  2. The Full Roster: 54 Agents and Their Superpowers
  3. Three-Minute Setup: Installing Every Agent
  4. Four Ways to Ask for Help (No Memorization Required)
  5. End-to-End Walk-Through: From Idea to Production in One Command
  6. Rolling Your Own Agent When the Built-ins Aren’t Enough
  7. Quick-Reference FAQ
  8. Decision Matrix: Which Agent Should I Call?

1. What Exactly Is Claude Code Unified Agents?

Imagine walking into a room where 54 seasoned specialists are on standby—each ready to jump in the moment you say,
“Help me design a REST API,”
or
“Set up a HIPAA-compliant patient portal,”
or even
“Explain why my Kubernetes rollout keeps failing.”

Claude Code Unified Agents turns that imaginary room into a real folder on your computer.

  • Every agent is a plain-text file you can read, edit, or fork.
  • All 54 agents ship with more than 1,000 lines of production-grade examples, comments, and best-practice notes.
  • You can summon one specialist or let a built-in orchestrator coordinate an entire team.

No external services, no hidden telemetry, no black-box binaries—just the knowledge you need, exactly where you need it.

2. The Full Roster: 54 Agents and Their Superpowers

Below are the agents exactly as they appear in the repository, grouped by domain.
For each entry you’ll find the agent name, a plain-English description, and the keywords that trigger it automatically.

🔧 Development (14 agents)

Agent What It Does Typical Keywords
backend-architect Designs REST or GraphQL APIs, picks the right database, splits monoliths into microservices “REST API design”, “microservice”, “PostgreSQL schema”
frontend-specialist Builds responsive UIs in React, Vue, Angular, or plain JavaScript “React component”, “Vue 3”, “responsive dashboard”
python-pro Writes idiomatic Python, asyncio, performance tuning, packaging “FastAPI”, “asyncio”, “Python optimization”
fullstack-engineer End-to-end apps: Node + React, Django + Vue, etc. “MERN stack”, “Django REST + React”
mobile-developer iOS, Android, React Native, Flutter codebases “Flutter login”, “Swift UI”
blockchain-developer Smart contracts, DeFi protocols, Web3 integrations “Solidity ERC-20”, “Hardhat deploy”
database-specialist SQL tuning, NoSQL modeling, migrations, backup plans “PostgreSQL index”, “MongoDB aggregation”
rust-pro Systems code, WebAssembly, memory-safe crates “Rust CLI”, “WASM module”
golang-pro Concurrent servers, cloud-native services “Go microservice”, “Gin API”
typescript-pro Large-scale TS, strict type configs “TypeScript monorepo”
javascript-pro Modern ES6+, Node.js tooling “Node script”, “ESLint setup”
java-enterprise Spring Boot, JVM tuning, microservices “Spring Security”, “JVM GC tuning”
nextjs-pro Next.js 14+, App Router, React Server Components “Next.js page”, “RSC fetch”
react-pro Advanced hooks, state libraries, performance tricks “React context”, “Zustand store”
vue-specialist Vue 3, Composition API, Nuxt 3, Pinia “Nuxt plugin”, “Pinia store”
angular-expert Angular 17+, signals, RxJS, enterprise patterns “Angular guard”, “RxJS operator”

🏗️ Infrastructure (7 agents)

Agent What It Does Typical Keywords
devops-engineer CI/CD pipelines, Docker, Kubernetes, GitHub Actions “GitHub Actions deploy”, “Dockerfile best practice”
cloud-architect Multi-cloud blueprints, cost estimates, IAM “AWS VPC”, “GCP budget alert”
incident-responder Reads logs, finds root cause, drafts runbooks “500 error spike”, “ELK query”
performance-engineer Profiling, load testing, optimization plans “p99 latency”, “JMeter script”
monitoring-specialist Metrics, alerts, dashboards (Prometheus, Grafana, DataDog) “Prometheus rule”, “SLI/SLO”
deployment-manager Release orchestration, blue-green, rollback playbooks “Canary deploy”, “helm rollback”
kubernetes-expert Helm charts, operators, CRDs, cluster hardening “K8s ingress”, “OPA policy”

✅ Quality (6 agents)

Agent What It Does Typical Keywords
code-reviewer Style, security, maintainability feedback “Code review”, “SonarQube issue”
security-auditor Vulnerability scans, OWASP checks “SQL injection”, “CVE check”
test-engineer Unit, integration, property-based test plans “pytest fixture”, “Jest mock”
e2e-test-specialist Playwright, Cypress, test strategies “Playwright test”, “Cypress CI”
performance-tester Load, stress, soak tests “k6 script”, “Locust swarm”
accessibility-auditor WCAG 2.2, screen-reader, keyboard nav “a11y check”, “axe-core”
performance-optimizer Bottleneck fixes, caching layers “Redis cache”, “CDN setup”

🤖 Data & AI (6 agents)

Agent What It Does Typical Keywords
ai-engineer LLM fine-tuning, vision pipelines, NLP “OpenAI fine-tune”, “Hugging Face model”
data-engineer ETL, Airflow DAGs, warehouse modeling “dbt model”, “Spark job”
data-scientist Stats, notebooks, predictive models “Pandas pipeline”, “scikit-learn”
mlops-engineer CI/CD for ML, model registry, drift monitoring “MLflow tracking”, “SageMaker endpoint”
prompt-engineer Prompt templates, RAG retrieval, token limits “RAG chain”, “prompt template”
analytics-engineer BI tools, metric definitions, SQL generators “LookML view”, “Metabase dashboard”

💼 Business & Process (8 agents)

Agent What It Does Typical Keywords
project-manager Sprint plans, story points, risk boards “Jira epic”, “burn-down chart”
product-strategist Market sizing, roadmap, KPI trees “MVP scope”, “OKR”
business-analyst Gap analysis, ROI models, process maps “BPMN diagram”, “cost-benefit”
technical-writer Docs, READMEs, API specs “OpenAPI spec”, “MkDocs site”
requirements-analyst User stories, traceability matrix “user story”, “acceptance criteria”
api-designer REST conventions, GraphQL schema, SDK stubs “OpenAPI yaml”, “GraphQL resolver”

🎨 Creative (3 agents)

Agent What It Does Typical Keywords
ux-designer Wireframes, user flows, personas “Figma prototype”, “user journey”
ui-designer Design systems, component libraries “Storybook”, “Tailwind theme”
content-strategist Editorial calendar, SEO outline “content brief”, “keyword cluster”

🧠 Meta-Management (5 agents)

Agent What It Does Typical Keywords
context-manager Restores sessions, compresses memory “restore context”, “trim tokens”
workflow-optimizer Faster CI, parallel builds, cache tuning “speed up GitHub Actions”
agent-generator Creates new agents from templates “add custom agent”
error-detective Stack-trace forensics, pattern matching “segfault cause”, “regex search”
documentation-writer Multi-format docs, auto-generated README “generate docs”

🔮 Specialized Domains (10 agents)

Agent What It Does Typical Keywords
blockchain-developer Web3 front-end, wallet connect, DeFi math “wagmi hook”, “Uniswap v3”
mobile-developer Native or cross-platform apps, store releases “App Store release”, “Android keystore”
game-developer Unity, Unreal Engine 5, Godot 4, procedural gen “Unity shader”, “UE5 Niagara”
iot-engineer MQTT, firmware OTA, edge analytics “ESP32 firmware”, “LoRaWAN”
embedded-engineer Arduino, STM32, RTOS, real-time constraints “FreeRTOS task”, “STM32CubeMX”
fintech-specialist PCI-DSS flows, payment gateways, fraud rules “Stripe webhook”, “PCI SAQ-A”
healthcare-dev FHIR resources, HIPAA checklists, EHR links “FHIR Patient”, “HIPAA BAA”
ecommerce-expert Cart, checkout, inventory, promotions engine “Shopify function”, “promotion rule”

3. Three-Minute Setup: Installing Every Agent

Option A: One-Line Script (Recommended)

git clone https://github.com/stretchcloud/claude-code-unified-agents.git
cd claude-code-unified-agents
./scripts/install.sh                # all 54 agents
# or cherry-pick
./scripts/install.sh --agents backend-architect,frontend-specialist

Option B: Project-Scoped Install

mkdir -p .claude/agents
cp -r claude-code-unified-agents/.claude/agents/* .claude/agents/

The agents live only inside your repo—perfect for open-source projects where contributors get the same helpers.

Option C: Global Install

mkdir -p ~/.claude/agents
cp -r claude-code-unified-agents/.claude/agents/* ~/.claude/agents/

Now every folder on your machine can call any agent without extra steps.

4. Four Ways to Ask for Help (No Memorization Required)

Method How You Use It Example
Interactive Browser Type /agents inside Claude Code A pick-list pops up; arrow keys to choose.
Auto-Detection Just describe the problem “Add JWT auth to my Express API” → backend-architect jumps in.
Direct Mention Prefix with @ @vue-specialist scaffold a dashboard with Pinia stores
Orchestrator Ask the conductor @orchestrator Build a HIPAA-compliant tele-health MVP

5. End-to-End Walk-Through: From Idea to Production in One Command

Goal
Launch a minimal yet production-ready e-commerce site with payments, inventory, and an admin panel—in under three days.

Step 1: One Sentence to the Orchestrator

@orchestrator Create a small e-commerce site that sells T-shirts, uses Stripe for payments, and keeps stock in Postgres.

Step 2: Watch the Plan Unfold

The orchestrator silently chats with eight agents:

graph TD
    A[You: idea] --> B[product-strategist: scope & MVP]
    B --> C[requirements-analyst: user stories]
    C --> D[backend-architect: API & DB schema]
    C --> E[frontend-specialist: React storefront]
    D --> F[database-specialist: indexes & migrations]
    E --> G[ui-designer: admin dashboard wireframes]
    D --> H[security-auditor: PCI-DSS checklist]
    H --> I[devops-engineer: GitHub Actions + AWS ECS]

Step 3: Day-by-Day Reality Check

Day Output You Receive Who Did It
1 openapi.yaml + ERD diagram + Stripe webhook plan backend-architect, database-specialist
2 React storefront + admin dashboard + seeded Postgres frontend-specialist, ui-designer
3 CI/CD pipeline + load-test report + PCI checklist devops-engineer, security-auditor

You review, tweak, and deploy—no Slack threads, no context-switching.

6. Rolling Your Own Agent When the Built-ins Aren’t Enough

Scenario
Your company uses an internal framework called “AcmeStack.” None of the 54 agents know it.

6.1 Create the Agent File

touch .claude/agents/development/acmestack-guru.md

6.2 Fill in the Template

---
name: acmestack-guru
description: Use when working with AcmeStack, our in-house framework
category: development
tools: Write, Read, MultiEdit, Bash
---

You are a senior engineer with 5 years of hands-on AcmeStack experience.
Always:
- Scaffold new services with `acme create-service`
- Use the built-in logger at `acme.log()`
- Run tests via `acme test --watch`

6.3 Use It Immediately

@acmestack-guru generate a user-service with CRUD endpoints

The new agent follows the same rules, shows up in /agents, and can even be added to the global install.

7. Quick-Reference FAQ

Q1: Does the install bloat my repo?
A: No. All agents together weigh under 5 MB—plain Markdown and snippets, no binaries.

Q2: Can I restrict an agent’s file-system access?
A: Yes. Edit the tools: line in the agent header. Example:

tools: Read, Write   # no Bash or Grep

Q3: My project already has a .claude folder. Will installing overwrite it?
A: Files are only added; nothing is deleted unless the names clash. Backups are automatic.

Q4: Do the agents send code to the cloud?
A: No. Everything runs locally inside Claude Code. No telemetry leaves your machine.

Q5: How do I update to the latest agents?

git pull origin main
./scripts/install.sh --update

8. Decision Matrix: Which Agent Should I Call?

If You Need … Call This Agent
A clean REST API spec @backend-architect
React performance fixes @frontend-specialist
Faster GitHub Actions @workflow-optimizer
HIPAA checklist for a health app @healthcare-dev
Unity multiplayer code @game-developer
Postgres slow-query fix @database-specialist
PCI-DSS payment flow @fintech-specialist
Storybook design-system setup @ui-designer
A/B test plan @analytics-engineer
Restore yesterday’s session @context-manager
Decode a weird segfault @error-detective

Closing Thoughts

The 54 agents are not a flashy plugin—they are a curated library of real-world solutions distilled from open-source repositories, conference talks, and on-call war stories.
Next time a task feels bigger than one brain, remember: you have 54 extra brains on speed dial.
Open your terminal and type:

claude "@orchestrator help me ship faster, safer, and smarter."

Then watch the room of experts go to work.