112 AI Agent Personas for Your IDE — We Tested All of Them
Agency-agents turns generic coding assistants into specialized experts with personality and rules
The Problem
Generic AI coding assistants give generic answers. Agency-agents provides 112 specialized personas — from Backend Architect to Security Engineer to Reality Checker — each with defined identity, rules, and success metrics. Works with Claude Code, Cursor, Copilot, and more. But do specialized personas actually produce better output than vanilla Claude?
Our Setup
We installed all 112 agents and tested the top 20 across real tasks: building a landing page, writing API endpoints, reviewing security, creating database schemas. Each task done twice — vanilla Claude vs agency-agents persona. Blind evaluation of output quality.
What We Expected
Specialized personas should outperform vanilla Claude on domain-specific tasks but make minimal difference on simple tasks. The real value should be consistency — same persona produces similarly structured output every time.
Want updates on this experiment?
Get notified when we publish new results or start related experiments.