Agency Agents: The Open-Source Project That Wants to Replace Your Entire Team

You hire a growth hacker. Then a backend developer. Then a QA engineer because your backend developer keeps shipping bugs. Then a project manager because nobody agrees on what gets built next. Then a paid media buyer because the growth hacker wants to test Google Ads. Then a UX researcher because the paid media buyer says conversions are tanking and nobody knows why.

Each of them costs money. Has opinions. Goes on holiday. Needs onboarding. Needs managing. Disagrees with the others. Gets blocked waiting on someone else.

Agency Agents starts from a different premise entirely. What if you just had all of them, always, ready to go, specialized, and free?

This is an open-source project built for Claude Code. 135 agents. 13 divisions. And it keeps growing.

What This Actually Is

This is not a chatbot. It is not a productivity layer on top of a generic AI. It is a full workforce blueprint.

Every agent in the library has a defined role, a professional personality, a domain of expertise, a set of deliverable standards, and a specific way of approaching work. You do not get a generic "help me code" assistant. You get a Frontend Developer who cares about accessibility and clean component architecture. A Growth Hacker who thinks about acquisition loops and retention metrics. A Paid Media Auditor who starts every engagement by tearing apart your tracking setup before touching a single campaign.

The difference is context. A blank AI assistant has none. These agents start with everything already loaded.

Here is what the full roster looks like across all 13 divisions:

Engineering Division (24 agents) covers the full technical stack. Frontend Developer, Backend Architect, AI Engineer, Security Engineer, Mobile App Builder, Site Reliability Engineer, Solidity Smart Contract Engineer, Git Workflow Master, and more. Real specialists with real domain depth, not one generic "developer" agent.

Marketing Division (23 agents) handles every channel. Growth Hacker, Content Creator, Twitter Engager, TikTok Strategist, Reddit Community Builder, LinkedIn Content Creator, SEO Specialist, Podcast Strategist. This is not a "write me a social post" tool. These agents have platform-specific context baked in.

Sales Division (8 agents) runs the revenue side. Outbound Strategist, Discovery Coach, Deal Strategist, Proposal Strategist, Pipeline Analyst. Think of it as a trained sales team that never forgets a framework and never has a bad day.

Design Division (8 agents) covers every surface. UI Designer, UX Researcher, Brand Guardian, Visual Storyteller, Inclusive Visuals Specialist. They do not just make things look nice. They reason about systems and constraints.

Paid Media Division (7 agents) manages the money you spend on traffic. PPC Campaign Strategist, Search Query Analyst, Paid Media Auditor, Ad Creative Strategist, Tracking Specialist. A full paid media team in one project.

Testing Division (8 agents) is the one most teams skip and later regret. Reality Checker, Evidence Collector, Performance Benchmarker, Accessibility Auditor. These agents exist to find the things you missed.

Product Division (4 agents) keeps the work pointed in the right direction. Sprint Prioritizer, Trend Researcher, Feedback Synthesizer. Strategy, not just execution.

Project Management Division (6 agents) handles the coordination layer. Project Shepherd, Senior Project Manager, Experiment Tracker, Studio Producer. The work that nobody wants to do but everything falls apart without.

Support Division (6 agents) runs the operational intelligence. Analytics Reporter, Finance Tracker, Legal Compliance Checker, Executive Summary Generator. Reporting, risk, and accountability.

Spatial Computing Division (6 agents) is where it gets genuinely futuristic. XR Interface Architect, visionOS Spatial Engineer, XR Cockpit Interaction Specialist. If you are building for Apple Vision Pro or any spatial computing platform, these agents already know the constraints.

Specialized Division (24 agents) is the catch-all for high-value niche expertise. Blockchain Security Auditor, Compliance Auditor, Recruitment Specialist, Supply Chain Strategist, Healthcare Marketing Compliance, MCP Builder, ZK Steward. Roles that would be prohibitively expensive to hire for. Available on demand.

That is 135 agents. You can use one at a time or deploy several simultaneously on the same project.

What This Looks Like in Practice

The best way to understand the scale of what is possible is through concrete scenarios. Here are five.

Scenario 1: Building a Startup MVP

You need a Frontend Developer to build the React app. A Backend Architect to design the API and database schema. A Growth Hacker planning the first 1000 users. A Rapid Prototyper cycling through ideas fast. A Reality Checker running quality gates before anything ships.

Five roles working in sequence. The Frontend Developer does not care about acquisition strategy. The Growth Hacker does not care about database design. Everyone is focused. You ship faster than a team of five humans because nobody argues about process and nobody is waiting on a standup to get unblocked.

Scenario 2: Marketing Campaign Launch

A Content Creator handles the copy. A Twitter Engager owns the platform strategy and scheduling. An Instagram Curator handles visual content. A Reddit Community Builder creates authentic presence in the communities where your actual audience lives, not just the ones where it feels safe to post.

An Analytics Reporter tracks what is working. Five specialists coordinating across channels simultaneously in a single session. No Slack messages. No missed handoffs.

Scenario 3: Enterprise Feature Development

A Senior Project Manager scopes the work and breaks it into tasks. A Senior Developer handles the implementation. A UI Designer builds out the design system components. An Experiment Tracker plans the A/B test framework. An Evidence Collector documents everything for the handoff. A Reality Checker signs off before it goes to production.

Six roles. One session. This is what enterprise delivery looks like without the enterprise overhead.

Scenario 4: Full Agency Product Discovery

This one is worth dwelling on because it shows the real ceiling.

Eight agents deployed simultaneously on a single product question. Product Trend Researcher, Backend Architect, Brand Guardian, Growth Hacker, Support Responder, UX Researcher, Project Shepherd, and XR Interface Architect all evaluate the same software opportunity and produce a unified product plan in one session.

Market validation, technical architecture, brand strategy, go-to-market planning, support systems design, UX research, project execution roadmap, and spatial UI design. One session. Eight specialists. This is the Nexus Spatial example from the Agency Agents documentation and it is the most compelling proof of scale the project has produced.

Scenario 5: Paid Media Account Takeover

A Paid Media Auditor does a full assessment of the account first. No assumptions. A Tracking Specialist verifies every conversion is firing correctly before anyone touches a budget. A PPC Campaign Strategist redesigns the account structure from scratch. A Search Query Analyst cleans up wasted spend in the search terms report. An Ad Creative Strategist refreshes all the copy and extensions. An Analytics Reporter builds the reporting dashboard that shows what is working going forward.

Systematic takeover. First 30 days. Tracking verified, waste eliminated, structure optimized. This is what a full-service paid media team delivers. Now you have it without the retainer.

How to Actually Use It

There are three ways in.

The native path is Claude Code. Agency Agents was built with Claude Code in mind, and the integration is direct. Each persona loads as a custom agent definition. Claude Code adopts the professional identity for the session. You switch roles as the work demands. This is the highest-fidelity version.

The portable path is prompt adaptation. The persona definitions can be extracted and used as system prompts for any capable LLM. Teams using GPT-4 or Gemini can run the same agents by adapting the prompt structure. The personas are tool-agnostic. What matters is the definition, not the interface.

The third path is conversion for coding tools. Cursor, Windsurf, Aider, and similar tools each accept persistent context in their own formats. The Agency Agents persona definitions convert cleanly into .cursorrules files, memory files, and system prompts for whichever development environment you prefer.

The Economic Argument

If you run a small agency or a startup, specialized talent is the constraint. You cannot afford a full-time paid media team and a QA division and a spatial computing specialist and a blockchain security auditor. The people who have that specialization are expensive. They are also hard to hire, hard to manage, and hard to replace.

Agency Agents makes the specialization free. The constraint shifts from budget to workflow discipline. The question is not whether you can afford a Security Engineer. The question is whether you will actually use one.

Teams that get the most out of this treat personas as workflow standards rather than optional experiments. Use the Frontend Developer agent for frontend work. Use the QA agent before every ship. Use the Paid Media Auditor before you touch a campaign budget. The consistency compounds.

This is not about replacing creativity or judgment. It is about not doing generic work when specialized work is available for free.

---

If you want to build with AI agent frameworks on your next project, come talk to us.