AI

Agency-Agents: Build Your AI Dream Team with 131k+ GitHub Stars

Explore Agency-Agents, the open-source collection of 230+ specialized AI personas with 131k+ GitHub stars. Learn how to build multi-agent workflows for any development role.

Tosin Akinosho

— 4 min read

Members-Only Deep Dive - This exclusive analysis is available to Decision Crafters community members.

Agency-Agents is an open-source collection of 230+ meticulously crafted AI agent personas designed to transform how teams build software. With 131k+ GitHub stars and active development (commits within the last 48 hours), this project has become the go-to resource for developers seeking structured, role-based AI agents that integrate seamlessly into modern development workflows. Whether you're building a startup MVP or orchestrating enterprise automation, Agency-Agents provides production-ready agent blueprints across 16 specialized divisions.

What is Agency-Agents?

Agency-Agents is a curated library of AI agent personas defined in markdown format. Created by Michael Sitarzewski and maintained by a growing community, the project provides a structured approach to multi-agent orchestration. Rather than building agents from scratch, developers can browse, customize, and deploy pre-designed personas that embody specific expertise—from frontend engineers to marketing strategists, from security architects to GIS specialists.

The project originated from a Reddit discussion and has evolved into a comprehensive ecosystem with 16 divisions spanning engineering, design, marketing, product, finance, healthcare, GIS, and more. Each agent includes detailed persona sections, technical deliverables, step-by-step workflows, success metrics, and real-world tested approaches. The repository is actively maintained with 386 commits and integrations for Claude Code, Cursor, Codex, Gemini, Qwen, and other AI platforms.

What makes Agency-Agents unique is its focus on structured persona design. Rather than generic prompts, each agent is a complete blueprint with identity, communication style, critical rules, learning mechanisms, and domain-specific expertise. This structured approach enables better consistency, reproducibility, and control when orchestrating multi-agent workflows.

Core Features and Architecture

230+ Specialized Agent Personas — The library includes agents across 16 divisions: Engineering (frontend, backend, mobile, AI, DevOps, security), Design (UI, UX, brand, interaction), Marketing (growth, content, social, paid media), Product (sprint planning, research, feedback), Project Management, Finance, Healthcare, GIS, Game Development, Spatial Computing, Testing, Support, Sales, and more. Each agent is production-ready and community-vetted.

Multi-Format Export — Agency-Agents supports 16 tool integrations including Claude Code, Cursor, Codex, Gemini, Qwen, OpenClaw, and others. The convert.sh script automatically transforms agent markdown into platform-specific formats, ensuring consistency across deployment targets. Generated output is gitignored, preventing build artifacts from polluting the repository.

Structured Persona Framework — Each agent follows a consistent template with frontmatter (name, emoji, vibe, services), Identity & Role Definition, Communication Style, Critical Rules, Technical Deliverables, Step-by-Step Workflows, Success Metrics, and Learning & Memory sections. This structure enables reliable parsing and consistent behavior across different AI platforms.

Division-Based Organization — Agents are organized into 16 divisions (stored in divisions.json), each with strategic, core, emerging, and delivery tiers. This hierarchical structure makes it easy to assemble teams for specific scenarios—startup MVPs, enterprise features, marketing campaigns, or incident response.

Tool Integration Ecosystem — The project includes integrations for 16 tools (tracked in tools.json). Each integration has its own converter, installer, and detector. Recent additions include ZCode (Z.ai GLM agent harness), demonstrating the project's commitment to supporting emerging AI platforms.

Community Contribution Pipeline — The CONTRIBUTING.md guide provides clear templates, originality checks, and linting rules. Pull requests are validated against duplicate detection, section header normalization, and tool integration checklists. This ensures quality and consistency as the library grows.

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

Prerequisites: Node.js 18+, bash, and git. The project is cross-platform (macOS, Linux, Windows via WSL2).

Installation: Clone the repository and run the install script for your target platform:

git clone https://github.com/msitarzewski/agency-agents.git
cd agency-agents
./scripts/install.sh --list teams  # View available divisions
./scripts/install.sh --tool claude-code  # Install for Claude Code

Browsing Agents: The web app (agency-agents-app) provides a searchable three-pane interface with division filters, category views, and per-agent detail panels. Alternatively, browse the GitHub repository directly—each agent is a markdown file in its division folder.

Customization: Agents are plain markdown files. Fork the repo, edit agent files, and run the linting tools to validate your changes. The originality checker prevents duplicates; the section header normalizer ensures consistent structure.

Real-World Use Cases

Startup MVP Assembly: Use the startup-mvp runbook to deploy a lean team: Product Manager, Frontend Engineer, Backend Engineer, DevOps Specialist, and QA Engineer. Each agent brings domain expertise and workflow consistency, reducing onboarding time for AI-assisted development.

Enterprise Feature Development: The enterprise-feature scenario orchestrates a full squad: Technical Lead, Architect, Frontend/Backend Engineers, Security Reviewer, and QA. This structured approach ensures security reviews, performance validation, and cross-functional alignment without manual coordination.

Marketing Campaign Execution: Deploy the marketing-campaign team (Growth Strategist, Content Creator, Social Media Manager, Paid Media Specialist, Analytics Lead) to coordinate campaign planning, content creation, channel management, and performance tracking—all with consistent brand voice and strategic alignment.

Incident Response Coordination: The incident-response runbook activates an on-call team (Incident Commander, Platform Engineer, Security Specialist, Communications Lead) to triage, remediate, and communicate during outages. Pre-defined workflows and escalation rules reduce MTTR and improve stakeholder communication.

How It Compares

vs. LangGraph/LangChain: LangGraph excels at low-level agent orchestration and state management. Agency-Agents is higher-level—it provides pre-built personas and workflows rather than primitives. Use LangGraph for custom agent logic; use Agency-Agents for rapid deployment of known roles.

vs. CrewAI: CrewAI focuses on multi-agent task orchestration with role-based agents. Agency-Agents is more comprehensive—it includes 230+ personas, tool integrations, and community-driven curation. CrewAI is better for custom orchestration; Agency-Agents is better for rapid team assembly.

vs. AutoGen: AutoGen (Microsoft) provides a framework for multi-agent conversations. Agency-Agents is a library of pre-built personas. AutoGen is more flexible for custom workflows; Agency-Agents is faster for known scenarios.

Strengths: Massive community library (230+ agents), multi-platform support (16 integrations), structured persona design, active maintenance, clear contribution guidelines.

Limitations: Requires manual customization for domain-specific needs, no built-in orchestration engine (relies on platform-specific implementations), learning curve for persona design best practices.

What is Next

The project roadmap includes expanding the agent library to 300+ personas, adding new divisions (e.g., legal, compliance, sustainability), improving the web app with real-time collaboration features, and building a community marketplace for agent templates. Recent additions like ZCode integration and healthcare division signal the project's commitment to emerging domains and platforms.

Agency-Agents represents a shift toward structured, reusable AI personas as a core development primitive. As multi-agent systems become standard in enterprise workflows, libraries like this will be essential for rapid, consistent team assembly. The project's 131k+ stars and active community suggest this is just the beginning.

Sources

Read more