Skip to main content

FrootAI — AmpliFAI your AI Ecosystem Get Started

FAI CLI

Scaffold projects, search knowledge, estimate costs, and validate configs — all from the terminal. Ships with the MCP Server package. No extra install needed.

No install needed — runs via npx

Terminal
npx frootai <command>

Or if you've installed frootai-mcp globally: just frootai <command>

Commands

frootai init

Interactive project scaffolding. 3 questions → complete FrootAI project with .github Agentic OS, DevKit, TuneKit, and MCP config.

frootai scaffold <play>

Scaffold a specific solution play (e.g., frootai scaffold 01-enterprise-rag). Creates the full project structure.

frootai search <query>

Search the FrootAI knowledge base from terminal. Returns relevant module sections.

frootai cost <play> [scale]

Estimate monthly Azure cost for a play. Usage: frootai cost 01 dev or frootai cost 07 prod.

frootai validate

Validate your project's FrootAI config files (.github structure, config/*.json, agent.md).

frootai doctor

Check your environment: Node version, MCP server status, VS Code extension, Git config.

frootai deploy <play>

Deploy a solution play to Azure. Walks through Bicep deployment with parameter prompts.

frootai help

Show all available commands and usage examples.

frootai mcp * — Federation Subcommands

v6.1.0 GA

Six subcommands to drive the FrootAI MCP federation router from the terminal. Attach any of the 7 Tier-1 MCP areas, invoke <area>.<tool> calls, query the trust manifest, or one-shot a tool call with auto-detach. State persists at ~/.frootai/mcp-state.json. Launch post →

frootai mcp list

List the 7 federated Tier-1 MCP areas (azure, github, playwright, markitdown, ms-learn, context7, chrome-devtools) + show currently-attached sessions

frootai mcp attach <area>

Attach a federated MCP area; runs trust gate, spawns stdio transport, returns the namespaced tool list (<area>.<tool>)

frootai mcp detach <area>

Detach + dispose an attached area; idle disconnect honored at the configured timeout

frootai mcp invoke <area>.<tool> [--args JSON]

Invoke a tool on an attached area. Returns headline + content blocks (text mode) or full JSON envelope (--json)

frootai mcp call <area>.<tool> [--args JSON]

One-shot: attach → invoke → detach in a single call (for ephemeral usage); always wraps in try/finally for clean dispose

frootai mcp trust <publisher>

Query the trust manifest for a publisher: returns tier (first-party-ms / verified / community / untrusted) + policy

Quick example
Terminal
$ frootai mcp list
7 areas available · 0 attached

  azure              first-party-ms  ✓ ready
  github             first-party-ms  ✓ ready
  playwright         first-party-ms  ✓ ready
  markitdown         first-party-ms  ✓ ready
  ms-learn           first-party-ms  ✓ ready
  context7           verified-publisher  ✓ ready
  chrome-devtools    verified-publisher  ✓ ready

$ frootai mcp attach azure
✓ trust gate: first-party-ms (auto-attach)
✓ spawned npx @azure/mcp-server (PID 12847, transport=stdio)
✓ registered 218 tools as azure.*

$ frootai mcp invoke azure.subscription_list
{
  "subscriptions": [
    { "id": "abc-123", "name": "Production", "state": "Enabled" },
    ...
  ]
}

$ frootai mcp call playwright.screenshot --args '{"url":"https://frootai.dev"}'
✓ attached playwright (cold) · invoked · detached
saved screenshot to ./playwright-screenshot.png (142 KB)

Example: frootai init

Terminal
$ npx frootai init

  🌳 FrootAI™ CLI v3.1.2
  From the Roots to the Fruits

  What are you building?
  1) Enterprise RAG
  2) AI Agent
  3) AI Gateway
  4) Content Moderation
  5) Multi-modal
  6) Custom (pick from all plays)

  Choose [1-6]: 1

  Target scale?
  1) dev   — Local development
  2) prod  — Production, HA

  Choose [1-2]: 1

  Project name [my-ai-project]: my-rag-app

  ✅ Created my-rag-app/ with:
  └── .github/ (Agentic OS — 19 files)
  └── .vscode/mcp.json
  └── config/ (openai.json, search.json, ...)
  └── infra/main.bicep
  └── evaluation/
  └── agent.md
MCP ServerSetup GuideEcosystemBack to FrootAI