AGENTS.md — FrootAI Cross-Platform Agent Discovery

Standard: This file follows the AGENTS.md specification  for cross-platform agent discovery. Any AI coding assistant (Copilot, Claude, Cursor, Windsurf, Cody, etc.) can read this file to discover available agents.

Overview

FrootAI provides specialized AI agents organized by domain. Each agent is a .agent.md file with YAML frontmatter defining its description, tools, model preferences, and WAF (Well-Architected Framework) alignment.

Browse the full catalog: frootai.dev/primitives/agents 

Agent Architecture

agents/
├── fai-rag-architect.agent.md              # RAG specialist
├── fai-azure-ai-search-expert.agent.md     # AI Search expert
├── fai-security-reviewer.agent.md          # Security reviewer
├── fai-play-01-builder.agent.md            # Enterprise RAG Builder
├── ... (agents)
└── fai-rag-architect/fai-context.json      # FAI Protocol context

Each agent follows this structure:

---
description: "One-line description of what this agent does"
tools: ["codebase", "terminal"]              # Tools the agent can use
model: ["gpt-4o", "gpt-4o-mini"]             # Preferred models (array)
waf: ["security", "reliability"]             # WAF pillar alignment
plays: ["01-enterprise-rag", "21-agentic-rag"] # Compatible solution plays
---
 
# Agent Name
 
Detailed instructions for the agent's behavior, expertise, and constraints.

Agent Categories

🔍 RAG & Search (15+ agents)

🤖 Agent & Multi-Agent (15+ agents)

🏗️ Infrastructure & Platform (20+ agents)

🔒 Security & Compliance (15+ agents)

🎙️ Voice & Speech (5+ agents)

📄 Document Processing (10+ agents)

🏥 Industry Specialists (20+ agents)

⚙️ DevOps & Tooling (15+ agents)

🧠 ML & Model Management (10+ agents)

🎨 Creative & Media (10+ agents)

🔌 Platform Agents (10+ agents)

🎯 Solution Play Agents (69 agents)

Each of the 23 solution plays has a dedicated builder/reviewer/tuner agent triad:

Plus 2 utility agents: FAI Play Dispatcher (routes to correct play) and FAI Play Lifecycle (play management).

How to Use

In VS Code (with GitHub Copilot)

# Copilot automatically discovers agents from .agent.md files
# Reference an agent in Copilot Chat:
@workspace Which agent should I use for RAG?

In a Solution Play

Agents are wired via fai-manifest.json:

{
  "primitives": {
    "agents": ["./agent.md", "../../agents/fai-rag-architect.agent.md"]
  }
}

Via MCP Server

npx frootai-mcp@latest  # 25 tools including agent_build, agent_review, agent_tune

Via CLI

npx frootai primitives --type agents  # Browse all agents

Related Resources

• Full Catalog: frootai.dev/primitives/agents 
• Agent Patterns (L3): frootai.dev/learning-hub/agent-patterns 
• Primitive Primer (L2): frootai.dev/learning-hub/primitive-primer 
• FAI Protocol: frootai.dev/fai-protocol 
• Solution Plays: frootai.dev/solution-plays 
• GitHub: github.com/frootai/frootai/tree/main/agents