Getting Started
FrootAI is an AI infrastructure LEGO kit. Pick one path, follow the steps, and your AI agent gets 45 tools + 100 solution plays + 860 primitives in under 5 minutes.
What is FrootAI?
Today, AI tools are disconnected: agents don't know about skills, skills don't reference config, hooks fire on every tool call. FrootAI fixes this with the FAI Protocol — a wiring layer that connects agents, skills, prompts, hooks, instructions, and config into one system.
Who are you?
Pick your role and we'll recommend the best install path:
Choose Your Install Path
MCP Server Setup
Works with Copilot, Claude, Cursor, Windsurf, Microsoft Foundry — any MCP client.
Prerequisites
Step 1 — Add MCP config to your project
Create .vscode/mcp.json in your workspace root:
{
"servers": {
"frootai": {
"type": "stdio",
"command": "npx",
"args": ["frootai-mcp@latest"]
}
}
}node -v — must be 18+. Then reload VS Code (Ctrl+Shift+P → Reload Window).▶ Alternative install methods (global npm, Docker, Git clone)
# Global install (available everywhere)
npm install -g frootai-mcp@latest
frootai-mcp
# Docker (no Node.js needed)
docker run -i --rm ghcr.io/frootai/frootai-mcp:latest
# From source
git clone https://github.com/frootai/frootai.git
cd frootai/mcp-server && npm installStep 2 — Connect your AI client
Pick your client:
Add to your config file (claude_desktop_config.json or .cursor/mcp.json):
{
"mcpServers": {
"frootai": {
"command": "npx",
"args": ["frootai-mcp"]
}
}
}Step 3 — Verify it works
Open your AI chat and try this prompt:
List all FrootAI solution plays@frootai to check if tools are visible. Reload the client if not.Step 4 — Initialize a DevKit (optional)
Ask your agent to scaffold a complete solution:
Initialize the Enterprise RAG DevKit in my workspaceThe agent calls get_play_detail, reads the manifest, and guides you through setting up all DevKit files.
How It Works (3 Steps)
Install
Add FrootAI via MCP, VS Code extension, npm, Docker, or pip
Init DevKit
Pick a solution play → agents, skills, prompts deploy into your workspace
Build with AI
Your AI agent discovers everything automatically. Start building.
What You Get
Every solution play ships a DevKit — a folder of files that wire your AI agent for a specific domain. When you initialize a play, these files land in your workspace:
.github/copilot-instructions.mdDomain knowledge auto-injected into every Copilot conversation
.github/agents/*.agent.md3 specialist agents: @builder, @reviewer, @tuner
.github/prompts/*.prompt.md4 slash commands: /deploy, /test, /review, /evaluate
.github/skills/*/SKILL.mdStep-by-step procedures (100+ lines each)
.github/hooks/guardrails.jsonSessionStart hook — sets context on workspace open
.github/instructions/*.mdCoding patterns injected when editing specific file types
config/*.jsonTunable parameters — model, temperature, thresholds
.vscode/mcp.jsonFrootAI MCP server — boots automatically
Try These Right Now
We have an AI assistant called — it knows every play, primitive, architecture pattern, and Azure pricing.
Click any prompt below and Agent FAI will answer right here — no install needed. Or copy-paste them into your own AI agent after setup.
1. Explore solution plays
Show me all FrootAI solution plays.
Which one is best for building a RAG pipeline?
Explain its architecture.2. Estimate Azure costs
Use FrootAI to estimate the monthly cost of Play 01
(Enterprise RAG) for dev, staging, and production.3. Learn best practices
What are the best practices for chunking documents
in a RAG system? Show the FrootAI architecture pattern.What's Next?
🚀 Build
Pick a play and initialize its DevKit.
📚 Learn
Guided paths from beginner to advanced.
🔍 Explore
Browse the full ecosystem.