The Agency: A Complete Guide to 232+ Specialized AI Agent Personas

What Is The Agency?

The Agency (github.com/msitarzewski/agency-agents) is an open-source collection of 232+ specialized AI agent personas — each with a distinct personality, domain expertise, and proven workflows. Born from a Reddit thread and months of real-world iteration, it's one of the most comprehensive open-source AI agent libraries available today.

What started as a Reddit post has evolved into a community-driven project with 10,000+ lines of agent definitions, battle-tested across production environments.

Why It Matters

Generic prompts like "act as a developer" only take you so far. The Agency gives you:

- Deep specialization — Each agent is a domain expert, not a generic template - Personality-driven — Unique voice, communication style, and approach for each agent - Deliverable-focused — Code examples, measurable outcomes, and proven workflows - Tool-agnostic — Works with Claude Code, Cursor, OpenClaw, GitHub Copilot, Windsurf, Codex, and 7+ other tools - Community-maintained — Active translations in Chinese, Portuguese, Japanese, Korean, Russian, Arabic, Indonesian

> "I don't just test your code — I default to finding 3-5 issues and require visual proof for everything." — Evidence Collector (Testing Division)

---

The 16 Divisions at a Glance

The agents are organized into divisions, each covering a major domain. Here's what you get:

---

Quick Start: Using The Agency with Any AI Tool

The Agency includes a powerful install script that supports 12 different AI coding tools. Here's how to get started:

Option 1: Install Everything (Recommended)

# Clone the repository
git clone https://github.com/msitarzewski/agency-agents.git
cd agency-agents
# Generate integration files for all supported tools
./scripts/convert.sh
# Run the interactive installer
./scripts/install.sh

The installer scans your system for installed tools and presents a checkbox UI:

  +------------------------------------------------+
  |   The Agency -- Tool Installer                 |
  +------------------------------------------------+
  System scan: [*] = detected on this machine
  [x]  1)  [*]  Claude Code     (claude.ai/code)
  [x]  2)  [*]  Copilot         (~/.github + ~/.copilot)
  [x]  3)  [*]  Antigravity     (~/.gemini/antigravity)
  [ ]  4)  [ ]  Gemini CLI      (~/.gemini/agents)
  [ ]  5)  [ ]  OpenCode        (opencode.ai)
  [ ]  6)  [ ]  OpenClaw        (~/.openclaw/agency-agents)
  [x]  7)  [*]  Cursor          (.cursor/rules)
  [ ]  8)  [ ]  Aider           (CONVENTIONS.md)
  [ ]  9)  [ ]  Windsurf        (.windsurfrules)
  [ ] 10)  [ ]  Qwen Code       (~/.qwen/agents)
  [ ] 11)  [ ]  Kimi Code       (~/.config/kimi/agents)
  [ ] 12)  [ ]  Codex           (~/.codex/agents)
  [1-12] toggle   [a] all   [n] none   [d] detected
  [Enter] install   [q] quit

Option 2: Target a Specific Tool

# Claude Code (native support — no conversion needed)
./scripts/install.sh --tool claude-code
# OpenClaw (converts agents to workspace format)
./scripts/convert.sh --tool openclaw
./scripts/install.sh --tool openclaw
# Cursor (creates .mdc rule files per agent)
./scripts/install.sh --tool cursor
# GitHub Copilot (native .md support)
./scripts/install.sh --tool copilot

Option 3: Install Only Specific Teams

Don't want all 232 agents? Install only what you need:

# Interactive wizard — pick tools + teams
./scripts/install.sh
# Install only Engineering and Security for Claude Code
./scripts/install.sh --tool claude-code --division engineering,security
# Install specific agents for Cursor
./scripts/install.sh --tool cursor --agent frontend-developer,ui-designer
# Preview what would be installed
./scripts/install.sh --tool opencode --division engineering --dry-run
# See every team and agent count
./scripts/install.sh --list teams

---

Deep Dive: How Agents Work

Each agent file is a markdown document with a structured format that serves as both documentation and a system prompt. Here's the anatomy of an agent:

# Agent Name
Identity & Memory
- Role: Digital marketing expert focused on SEO and content strategy
- Voice: Direct, data-driven, and slightly irreverent
- Knowledge Cutoff: Your standard knowledge cutoff applies
Core Mission
Your mission is to drive sustainable organic search growth...
Critical Rules
1. Never recommend black-hat SEO techniques
2. Always provide reasoning for recommendations
3. Cite data sources when making claims
Workflow Process
When starting a new SEO engagement:
1. Perform a technical site audit
2. Analyze current rankings and competitor landscape
3. Prioritize findings by impact
Technical Deliverables
Example: SEO Audit Report

[Structured output with findings, priorities, and action items]

Success Metrics
- Organic traffic growth: +30% within 6 months
- Keyword ranking improvements tracked weekly

Real Agent Example: Frontend Developer

The Frontend Developer agent (Engineering Division) handles modern web apps, UI implementation, and Core Web Vitals optimization. Its includes:

- Identity: React/Vue/Angular specialist with strong opinions on performance - Critical Rules: No jQuery, always optimize for Core Web Vitals, accessibility is non-negotiable - Workflow: Design system first → component library → integration → performance audit - Success Metrics: 90+ Lighthouse score, zero accessibility violations, < 100KB JS bundle

---

Real-World Use Cases

Scenario 1: Building a Startup MVP

Assemble your dream team:

1. 🎨 Frontend Developer — Build the React app 2. 🏗️ Backend Architect — Design API and database schema 3. 🚀 Growth Hacker — Plan user acquisition strategy 4. ⚡ Rapid Prototyper — Fast iteration cycles 5. 🔍 Reality Checker — Quality gate before launch

Result: A production-ready MVP with specialized expertise at every stage.

Scenario 2: Paid Media Account Takeover

A systematic 30-day plan:

1. 📋 Paid Media Auditor — 200+ point account assessment 2. 📡 Tracking Specialist — Verify conversion tracking accuracy 3. 💰 PPC Strategist — Redesign account architecture 4. 🔍 Search Query Analyst — Clean up wasted spend 5. ✍️ Ad Creative Strategist — Refresh ad copy and extensions

Result: Systematic turnaround with tracking verified, waste eliminated, and structure optimized.

Scenario 3: Smart Campus Digital Twin

Geospatial + building data experts working together:

1. 🧠 GIS Technical Consultant — Define digital twin strategy 2. 🏗️ BIM/GIS Specialist — Convert Revit models to GIS scene layers 3. 🛸 Drone Mapping Specialist — Fly campus, generate orthomosaic and 3D mesh 4. 🌐 Web GIS Developer — Build the dashboard with MapLibre 5. 🏔️ 3D Scene Developer — Create immersive flyover tour

Result: A campus digital twin combining BIM detail, drone reality capture, and web accessibility.

Scenario 4: Enterprise Feature Development

With proper quality gates:

1. 👔 Senior Project Manager — Scope and task planning 2. 💎 Senior Developer — Complex implementation 3. 🎨 UI Designer — Design system and components 4. 🧪 Experiment Tracker — A/B test planning 5. 📸 Evidence Collector — Visual proof QA 6. 🔍 Reality Checker — Final production readiness sign-off

---

Multi-Tool Integration Details

The Agency's strength is its tool-agnostic design. Here's how each tool handles agents:

---

Performance: Parallel Install

On multi-core machines, use parallel mode for faster installs:

# Convert all integration files in parallel
./scripts/convert.sh --parallel
# Install all detected tools in parallel
./scripts/install.sh --no-interactive --parallel
# Control parallelism with --jobs
./scripts/convert.sh --parallel --jobs 8
./scripts/install.sh --no-interactive --parallel --jobs 4

---

Community Translations

The Agency has been localized into 8 languages by the community:

---

Customization: Creating Your Own Agent

Contributing is straightforward. Fork the repository and create a new .md file in the appropriate category:

---
name: "Your Agent Name"
description: "One-line description"
color: "#hexcolor"
---
Identity & Memory
Who is this agent? What's their voice?
Core Mission
What is this agent's primary purpose?
Critical Rules
1. Domain-specific constraints
2. Deal-breakers and non-negotiables
Technical Deliverables
Example: Concrete Output

Real examples that demonstrate the agent's work

Workflow Process
Step-by-step for common scenarios
Success Metrics
How is this agent's success measured?

Submit a PR — the maintainers are active and welcoming.

---

Pricing & License

Cost: $0. It's fully open source under the MIT License.

You can use The Agency commercially, personally, or in production. Attribution is appreciated but not required.

---

Summary

The Agency (agency-agents) is a remarkable open-source project that solves a real problem: generic AI prompts are insufficient for complex work. With 232+ specialized agents across 16 divisions, comprehensive multi-tool support, and an active community, it's one of the best resources available for anyone serious about leveraging AI agents in their workflow.

What sets it apart from other agent collections: - Not just prompts — deeply researched agent personas with real workflows - Works with 12 different coding tools out of the box - Battle-tested in production environments - Active community with translations in 8 languages - Comprehensive — covers everything from game dev to GIS to finance

Get started in 2 minutes:

git clone https://github.com/msitarzewski/agency-agents.git
cd agency-agents
./scripts/install.sh