Agency-Agents: Build Your Dream Team of AI Specialists with 63.9k+ GitHub Stars

Agency-Agents is an open-source framework created by Michael Sitarzewski that packages AI specialists as clearly defined, personality-driven agents. Born from a Reddit thread and months of real-world iteration, it provides 144 meticulously crafted AI agent personalities organized across 12 specialized divisions: Engineering, Design, Paid Media, Sales, Marketing, Product, Project Management, Testing, Support, Spatial Computing, Game Development, and Academic.

Unlike generic prompt templates or one-off AI tools, each agent comes with a distinct personality, proven workflows, technical deliverables with code examples, and measurable success metrics. The framework is designed to work with Claude Code, Cursor, Aider, Windsurf, Gemini CLI, OpenCode, and Kimi Code—making it compatible with the entire modern AI coding ecosystem.

The project has grown explosively, reaching 63.9k GitHub stars in just weeks, making it one of the fastest-growing open-source projects in 2026. It is actively maintained with commits happening daily and a thriving community contributing new agents and improvements.

What is Agency-Agents?

Agency-Agents solves a critical problem in AI-assisted development: generic AI prompts lack specialization, personality, and proven workflows. Instead of asking Claude or GPT to "act as a developer," you activate a specific agent with deep domain expertise, a distinct communication style, and battle-tested processes.

The framework emerged from a Reddit discussion about AI agent design patterns. Michael Sitarzewski iterated on the concept for months, testing agents in production environments before releasing the initial 51-agent collection. The community response was immediate and overwhelming—the project gained 10,000 stars in its first week and has continued accelerating.

Each agent is not a generic template. The Frontend Developer has specific opinions about React performance optimization and Core Web Vitals. The Reddit Community Builder focuses on becoming a valued member rather than pushing marketing messages. The Whimsy Injector adds playful elements that serve functional or emotional purposes. These are personalities with processes, not prompts with placeholders.

Core Features and Architecture

1. 144 Specialized AI Agents Across 12 Divisions

The Engineering Division includes 26 agents: Frontend Developer, Backend Architect, Mobile App Builder, AI Engineer, DevOps Automator, Security Engineer, Database Optimizer, Embedded Firmware Engineer, Solidity Smart Contract Engineer, and many more specialized roles.

The Marketing Division has 30 agents covering every channel and region: Growth Hacker, Content Creator, Twitter Engager, TikTok Strategist, Instagram Curator, Reddit Community Builder, SEO Specialist, Baidu SEO Specialist, Douyin Strategist, China Market Localization Strategist, and region-specific e-commerce operators.

The Design Division provides 8 agents: UI Designer, UX Researcher, UX Architect, Brand Guardian, Visual Storyteller, Whimsy Injector, Image Prompt Engineer, and Inclusive Visuals Specialist. The Sales Division has 8 agents focused on pipeline generation and deal closure. The Product Division coordinates 5 agents for discovery and prioritization.

Game Development is particularly comprehensive with 20+ agents across Unity, Unreal Engine, Godot, Blender, and Roblox Studio. The Academic Division includes Anthropologist, Geographer, Historian, Narratologist, and Psychologist for world-building and narrative design with scholarly rigor.

2. Personality-Driven Design

Each agent has a distinct personality and communication style. The Evidence Collector defaults to finding 3-5 issues and requires visual proof for everything. The Reality Checker certifies production readiness with evidence-based quality gates. The Whimsy Injector balances delight with functionality.

This personality-driven approach reduces hallucinations and improves output quality. When an agent has a specific identity and mission, it produces more consistent, focused results than generic prompts.

3. Production-Ready Workflows

Every agent includes step-by-step workflows, technical deliverables with code examples, success metrics, and communication styles. These are not vague guidelines—they are battle-tested processes from real-world usage.

4. Multi-Tool Integration

Run ./scripts/convert.sh to generate integration files for Cursor, GitHub Copilot, Aider, Windsurf, Gemini CLI, OpenCode, and Kimi Code. The ./scripts/install.sh script auto-detects your tools and installs agents interactively.

Get free AI agent insights weekly

Join our community of builders exploring the latest in AI agents, frameworks, and automation tools.

Join Free

Getting Started

Option 1: Use with Claude Code (Recommended)

The simplest path is with Claude Code:

cp -r agency-agents/* ~/.claude/agents/
# Now activate any agent in your Claude Code sessions

Option 2: Use as Reference

Browse the agents on GitHub and copy/adapt the ones you need. Each agent file contains identity traits, core mission, technical deliverables with code examples, and success metrics.

Option 3: Multi-Tool Installation

./scripts/convert.sh
./scripts/install.sh
./scripts/install.sh --tool cursor
./scripts/install.sh --tool copilot

Prerequisites

Git, Bash, one of the supported AI coding tools, and an LLM API key (Claude, GPT, Gemini, or compatible).

Real-World Use Cases

Scenario 1: Building a Startup MVP

Assemble Frontend Developer (React app), Backend Architect (API and database), Growth Hacker (user acquisition), Rapid Prototyper (fast iteration), and Reality Checker (quality gates). Each agent brings specialized expertise without hiring overhead. Result: Ship faster with specialized expertise at every stage.

Scenario 2: Marketing Campaign Launch

Deploy Content Creator (campaign content), Twitter Engager (Twitter strategy), Instagram Curator (visual content), Reddit Community Builder (authentic engagement), and Analytics Reporter (performance tracking). Each agent handles their platform with deep expertise. Result: Multi-channel coordinated campaign with platform-specific knowledge.

Scenario 3: Enterprise Feature Development

Coordinate Senior Project Manager (scoping), Senior Developer (complex implementation), UI Designer (design system), Experiment Tracker (A/B testing), Evidence Collector (QA), and Reality Checker (production readiness). Result: Enterprise-grade delivery with quality gates and documentation.

How It Compares

vs. Generic AI Prompts

Generic prompts lack specialization, personality, and proven workflows. Agency-Agents provides deep domain expertise with personality-driven approaches and measurable success metrics.

vs. Prompt Libraries

Prompt libraries offer one-off collections without integration or coordination. Agency-Agents provides comprehensive agent systems with workflows, deliverables, and multi-tool integration.

vs. Specialized AI Tools

Tools like Cursor or GitHub Copilot are black boxes you cannot customize. Agency-Agents is transparent, forkable, and adaptable—you own and control your agents.

Strengths

144 pre-built, battle-tested agents. Works with 8+ AI coding tools. MIT-licensed and community-driven. Personality-driven design reduces hallucinations. Proven workflows from real-world usage. Active maintenance and rapid growth.

Limitations

Requires manual setup and configuration. Depends on external LLM APIs. Learning curve for new users. Quality depends on the underlying LLM model.

What is Next

The community is actively adding new agents in emerging domains like blockchain security, healthcare compliance, and government digital transformation. Recent additions include Feishu integration, WeChat Mini Programs, and cross-border e-commerce specialists.

The project is exploring deeper integrations with MCP (Model Context Protocol) servers, enabling agents to access external tools and APIs seamlessly. There is discussion of building a visual agent marketplace and Agent of the Week showcase series.

Michael Sitarzewski has also released AGENT-ZERO, a complementary framework that gives teams using AI coding tools a shared contract, state machine, and documentation patterns to ship real software safely. This suggests the ecosystem is evolving toward production-grade agent orchestration.

As AI agents become the primary interface for software development and business operations, Agency-Agents is positioning itself as the canonical library of specialized agent personalities.

Sources

Agency-Agents GitHub Repository (March 27, 2026). Medium: Someone Built a Full AI Agency on GitHub (March 2026). LinkedIn: Fastest-Growing GitHub Projects (March 2026). AIBase News: Agency-Agents Surpasses 60,000 Stars (March 24, 2026). AGENT-ZERO: Shared Agent Contracts (March 2026).