FrootAI

_{MCP Server (Python)}

From the Roots to the Fruits. It's simply Frootful.

An open ecosystem where Infra, Platform, and App teams build AI — Frootfully.

A uniFAIng glue for the GenAI ecosystem, enabling deterministic and reliable AI solutions.

The Philosophy Behind FrootAI — The Essence of the FAI Engine

FrootAI is an intelligent way of packaging skills, knowledge, and the essential components of the GenAI ecosystem — all synced, not standalone. Infrastructure, platform, and application layers are woven together so that every piece understands and builds on the others. That's what "from the roots to the fruits" means: a fully connected ecosystem where Infra, Platform, and App teams build AI — Frootfully.

The FAI Ecosystem

Quick Start

Requirements: Python >= 3.10

pip install frootai-mcp

Run as MCP Server

frootai-mcp-py

Use in Python

from frootai_mcp.server import mcp

# Run as MCP server (stdio transport)
mcp.run(transport="stdio")

Or call tools directly:

import asyncio
from frootai_mcp.server import search_knowledge, wire_play

result = asyncio.run(search_knowledge(query="RAG architecture"))
print(result)

Connect to Your Agent

VS Code / GitHub Copilot .vscode/mcp.json:

{
  "servers": {
    "frootai": {
      "type": "stdio",
      "command": "frootai-mcp-py"
    }
  }
}

{
  "mcpServers": {
    "frootai": {
      "command": "frootai-mcp-py"
    }
  }
}

MCP Capabilities

MCP Tools (62)

Knowledge (6) — bundled knowledge, works offline

• list_modules — browse FROOT knowledge modules by layer
• get_module — read any module in full
• lookup_term — AI/ML glossary lookup (comprehensive glossary)
• search_knowledge — BM25 full-text search across all modules
• get_architecture_pattern — architecture decision guides
• get_froot_overview — complete framework summary

Solution Plays (5) — 101 pre-architected solutions

• list_solution_plays — list all 101 plays with filters
• get_play_detail — full play info with infra, tuning, complexity
• semantic_search_plays — BM25-powered play matching
• compare_plays — side-by-side play comparison
• generate_architecture_diagram — Mermaid.js diagrams

Agent Chain (3) — build → review → tune

• agent_build — architecture guidance + recommended play
• agent_review — security, quality, compliance audit
• agent_tune — production readiness + tuning recommendations

Azure / Live (4) — Azure AI intelligence

• get_model_catalog — model catalog with pricing tiers
• get_azure_pricing — service pricing by tier
• compare_models — side-by-side model comparison
• estimate_cost — itemized monthly cost per play

Ecosystem (5) — platform awareness

• get_github_agentic_os — .github Agentic OS guide
• list_community_plays — community plugin marketplace
• fetch_azure_docs — Azure documentation links
• fetch_external_mcp — discover MCP servers
• get_play_spec — SpecKit with WAF alignment

FAI Engine (5) — protocol wiring

• wire_play — generate fai-manifest.json for a play
• inspect_wiring — check what primitives are connected
• validate_manifest — validate fai-manifest.json schema
• validate_config — validate AI config parameters
• evaluate_quality — run quality evaluation (groundedness, relevance, etc.)

Live Evaluation (1) — v6.6+

• run_eval_live — POST test cases to a live endpoint, score with simple/LLM-as-judge evaluator, return aggregate mean/p95 vs threshold + per-case rows

Prompt Iteration (1) — v6.8.1+

• prompt_dry_run — test a system prompt with a single sample query without deploying. Calls Azure OpenAI when AZURE_OPENAI_ENDPOINT is set, else uses a deterministic shim (zero tokens). Returns the response + lightweight self-checks (citations, abstention, refusal, length, optional groundedness hint via lexical overlap with context/groundTruth) + iteration suggestions.

Marketplace (4) — primitive discovery

• list_marketplace — browse AI primitives by type
• get_primitive_detail — get detail for a specific primitive
• search_marketplace — keyword search across all primitives
• embedding_playground — v6.7+ similarity comparison with auto-selected backend (azure-openai 1536-dim when env set, else char-ngram TF-IDF cosine, with jaccard back-compat)

Scaffold (5) — project bootstrapping

• scaffold_play — scaffold a new play with DevKit structure
• smart_scaffold — describe what you want, get the best play
• list_templates — available scaffold templates by complexity
• preview_scaffold — dry-run preview of scaffold output
• scaffold_status — check scaffold completeness

Component & Config Scaffolding (3) — v6.4+

• scaffold_component — copy-paste-ready code for a single Azure AI building block (15 components: speech, search, openai, content-safety, key vault, cosmos, event hub, etc.)
• get_play_config — starter configs for a play (openai.json, search.json, speech.json, call-flow.json, etc.)
• get_dependencies — pinned dependency lists per language for a solution play

IaC Generation (1) — v6.5+

• generate_bicep — play-specific Bicep IaC with Managed Identity, RBAC, tags, optional private endpoints (plays 01, 03, 04, 06, 09, 14, 17, 21)

Workspace (1) — v6.2+

• analyze_workspace — inspect cwd: detected play, primitives, configs, Bicep services, FAI manifest, gaps + recommendations (sandboxed)

Extra (8) — specialized utilities

• run_evaluation — run evaluation with custom thresholds
• get_bicep_best_practices — Bicep IaC best practices
• list_primitives — list primitives by type
• get_waf_guidance — WAF pillar guidance
• check_play_compatibility — check if plays can compose
• get_learning_path — curated learning paths by topic
• export_play_config — export play config as JSON
• get_version_info — server version and capabilities

Runtime & Federation (10) — lean delivery, trust, and external MCP routing

• fai_lean / fai_full — switch between compact and full-fidelity content
• marketplace_spec / trust_evidence — inspect bundled attach specs and publisher evidence
• fai_attach_mcp / fai_detach_mcp — attach and release trust-gated MCP areas
• fai_list_attached / fai_discover_mcp — inspect active areas and discover compatible servers
• fai_trust_query / fai_invoke_via — evaluate trust and route a tool call through an area

MCP Resources (4)

MCP Prompts (6)

What Ships Inside

Feature parity with the Node.js MCP server — same MCP tools, same knowledge, same FAI Engine.

Testing

pip install pytest
cd python-mcp
python -m pytest tests/ -v
# 516 non-live tests (live provider suites are opt-in)

Links

© 2026 FrootAI — MIT License

_{AI architecture · MCP · model-context-protocol · Python · Azure · RAG · agents · copilot · semantic-kernel · open-source · frootai}