Complete Feature Specification
Every feature, tool, command, module, and integration — documented in one place.
FrootAI is a full-stack AI solution platform that accelerates Azure AI development. It provides pre-tuned solution plays, developer tooling (MCP + VS Code), a curated knowledge base, and infrastructure-as-code — all connected through the .github Agentic OS.
| Feature | Description | Status | Link |
|---|---|---|---|
| Website | 15-page Docusaurus site (dark theme, responsive, SEO-optimized) | ✅ Shipped | Home → |
| npm Package | [email protected] — 16 MCP tools (6 static + 4 live + 3 chain + 3 AI ecosystem) | ✅ Shipped | MCP Tooling → |
| VS Code Extension | [email protected] — 13 commands, standalone engine, sidebar navigation | ✅ Shipped | Extension → |
| GitHub Repo | Public repository, MIT license, CI/CD pipeline, 380+ files | ✅ Shipped | GitHub → |
| Azure Integration | AI Foundry-connected, Managed Identity, Bicep IaC, private endpoints | ✅ Shipped | Admin Guide → |
| Knowledge Base | 664KB curated content — 18 modules across 5 FROOT layers | ✅ Shipped | Packages → |
| Solution Framework | 20 pre-tuned solution plays with DevKit + TuneKit + Evaluation | ✅ Shipped | Plays → |
| AI Assistant | Chatbot for play recommendation + cost estimation (preview) | 🔄 Preview | Chatbot → |
| Plugin Marketplace | Decentralized marketplace for community agents, skills, prompts | 🔄 Preview | Marketplace → |
Each play ships with the full .github Agentic OS (19 files, 4 layers), DevKit (empower your coding agent), TuneKit (fine-tune AI for production), infrastructure blueprints, and evaluation scripts. LEGO blocks that compose into complete solutions.
| # | Play Name | Complexity | Status | Link |
|---|---|---|---|---|
| 01 | 🔍 Enterprise RAG Q&A | Medium | ✅ Ready | User Guide → |
| 02 | ⛰️ AI Landing Zone | Foundation | ✅ Ready | User Guide → |
| 03 | 🎯 Deterministic Agent | Medium | ✅ Ready | User Guide → |
| 04 | 📞 Call Center Voice AI | High | ✅ Ready | User Guide → |
| 05 | 🎫 IT Ticket Resolution | Medium | ✅ Ready | User Guide → |
| 06 | 📄 Document Intelligence | Medium | ✅ Ready | User Guide → |
| 07 | 🤖 Multi-Agent Service | High | ✅ Ready | User Guide → |
| 08 | 💬 Copilot Studio Bot | Low | ✅ Ready | User Guide → |
| 09 | 🔎 AI Search Portal | Medium | ✅ Ready | User Guide → |
| 10 | 🛡️ Content Moderation | Low | ✅ Ready | User Guide → |
| 11 | 🏔️ Landing Zone Advanced | High | ✅ Ready | User Guide → |
| 12 | ⚙️ Model Serving AKS | High | ✅ Ready | User Guide → |
| 13 | 🔬 Fine-Tuning Workflow | High | ✅ Ready | User Guide → |
| 14 | 🚪 AI Gateway (APIM) | Medium | ✅ Ready | User Guide → |
| 15 | 🖼️ Multi-Modal DocProc | Medium | ✅ Ready | User Guide → |
| 16 | 👥 Copilot Teams Extension | Medium | ✅ Ready | User Guide → |
| 17 | 📊 AI Observability | Medium | ✅ Ready | User Guide → |
| 18 | 📝 Prompt Management | Medium | ✅ Ready | User Guide → |
| 19 | 📱 Edge AI Phi-4 | High | ✅ Ready | User Guide → |
| 20 | 🚨 Anomaly Detection | High | ✅ Ready | User Guide → |
The .github folder evolved into a full agentic operating system. 7 primitives across 4 layers give your coding agent solution-aware context, guardrails, and chained workflows — before you write a single line of code.
| Feature | Description | Status | Link |
|---|---|---|---|
| copilot-instructions.md | Layer 1: Always-on solution context — Copilot reads this on every request | ✅ Shipped | Admin Guide → |
| instructions/*.instructions.md | Layer 1: Modular instruction files — azure-coding, security, patterns, testing | ✅ Shipped | API Ref → |
| prompts/*.prompt.md | Layer 2: 4 slash commands — /deploy, /test, /review, /evaluate | ✅ Shipped | Extension → |
| agents/*.agent.md | Layer 2: 3 chained specialists — builder → reviewer → tuner (auto-chain) | ✅ Shipped | Architecture → |
| skills/*/SKILL.md | Layer 3: 3 skill folders — deploy-azure, evaluate, tune (deep expertise) | ✅ Shipped | Admin Guide → |
| hooks/guardrails.json | Layer 4: Lifecycle enforcement — preToolUse policy gates | ✅ Shipped | API Ref → |
| workflows/*.md | Layer 4: Agentic CI/CD — AI-driven build, test, deploy workflows | ✅ Shipped | Architecture → |
| infra/main.bicep | Azure infrastructure — real Bicep resources (AI Foundry, Search, Container Apps) | ✅ Shipped | Setup → |
| agent.md | Rich play-specific personality (1500+ bytes) — shapes co-coder behavior | ✅ Shipped | User Guide → |
| plugin.json | Layer 4: Distribution manifest — marketplace packaging and discovery | ✅ Shipped | Marketplace → |
Ctrl+Shift+P → FrootAI: Initialize DevKit → Select a solution play → FrootAI copies the full .github Agentic OS (19 files) to your workspace. Copilot immediately becomes solution-aware.TuneKit provides pre-tuned configuration files so you can adjust AI behavior (temperature, top-k, models, guardrails) without being an AI specialist. Every parameter has been calibrated per solution play.
| Feature | Description | Status | Link |
|---|---|---|---|
| config/openai.json | Model settings — temperature, top-p, max-tokens, model selection, frequency penalty | ✅ Shipped | API Ref → |
| config/search.json | Search tuning — hybrid weights, semantic config, top-k, reranking, relevance threshold | ✅ Shipped | API Ref → |
| config/guardrails.json | Safety rails — content safety levels, blocklists, PII detection, topic restrictions | ✅ Shipped | API Ref → |
| config/chunking.json | Document chunking — chunk size, overlap, strategy (fixed/semantic/recursive) | ✅ Shipped | User Guide → |
| config/routing.json | Agent routing — model fallback chains, load balancing, priority routing | ✅ Shipped | API Ref → |
| config/agents.json | Agent behavior — personality, tools, memory, handoff rules, chain config | ✅ Shipped | API Ref → |
| config/model-comparison.json | Cost vs quality matrix — compare models across latency, price, quality dimensions | ✅ Shipped | MCP Tools → |
| evaluation/ | Evaluation scripts — eval.py, test datasets, quality targets, scoring metrics | ✅ Shipped | API Ref → |
| infra/main.bicep | Infrastructure config — SKUs, regions, scaling rules, private endpoints, RBAC | ✅ Shipped | Setup → |
The FrootAI MCP Server ([email protected]) gives your AI agent direct access to curated knowledge, architecture patterns, model catalog, and pricing data. Works with Copilot, Claude, Cursor, Gemini, and any MCP-compatible client.
| Feature | Description | Status | Link |
|---|---|---|---|
| list_modules | Browse all 18 FROOT modules organized by layer (F/R/O/O/T) | ✅ Shipped | MCP → |
| get_module | Read any module by ID (F1–T3) with optional section filtering | ✅ Shipped | MCP → |
| lookup_term | 200+ AI/ML term definitions — precise, curated glossary | ✅ Shipped | MCP → |
| search_knowledge | Full-text search across all 18 modules — ranked results | ✅ Shipped | MCP → |
| get_architecture_pattern | 7 pre-built decision guides (RAG, agents, hosting, cost, etc.) | ✅ Shipped | MCP → |
| get_froot_overview | Complete FROOT framework summary — layers, philosophy, structure | ✅ Shipped | MCP → |
| Feature | Description | Status | Link |
|---|---|---|---|
| fetch_azure_docs | Search Microsoft Learn for Azure documentation in real-time | ✅ Shipped | MCP → |
| fetch_external_mcp | Find MCP servers from community registries | ✅ Shipped | MCP → |
| list_community_plays | List all 20 solution plays from GitHub (live metadata) | ✅ Shipped | MCP → |
| get_github_agentic_os | .github 7 primitives guide — the agentic OS reference | ✅ Shipped | MCP → |
| Feature | Description | Status | Link |
|---|---|---|---|
| agent_build | Builder agent — architecture guidance, code scaffolding, suggests review | ✅ Shipped | MCP → |
| agent_review | Reviewer agent — security + quality checklist, suggests tune | ✅ Shipped | MCP → |
| agent_tune | Tuner agent — production readiness verdict, config optimization | ✅ Shipped | MCP → |
| Feature | Description | Status | Link |
|---|---|---|---|
| get_model_catalog | Azure AI model catalog — pricing, capabilities, context windows, recommendations | ✅ Shipped | MCP → |
| get_azure_pricing | Monthly cost estimates for RAG, agent, batch, realtime scenarios (dev/staging/prod) | ✅ Shipped | MCP → |
| compare_models | Side-by-side model comparison — recommend best model for use case + priority | ✅ Shipped | MCP → |
The FrootAI VS Code Extension ([email protected]) puts solution plays, AI modules, and MCP tools directly in your editor sidebar. 13 commands accessible via Ctrl+Shift+P.
| Feature | Description | Status | Link |
|---|---|---|---|
| FrootAI: Initialize DevKit | Full .github Agentic OS (19 files) + agent.md + MCP config + plugin.json | ✅ Shipped | Extension → |
| FrootAI: Initialize TuneKit | config/*.json + infra/main.bicep + evaluation/ — AI tuning for production | ✅ Shipped | Extension → |
| FrootAI: Install MCP Server | Install globally, run via npx, or add .vscode/mcp.json config | ✅ Shipped | Extension → |
| FrootAI: Start MCP Server | Launch frootai-mcp in terminal (16 tools: 6 static + 4 live + 3 chain + 3 AI) | ✅ Shipped | Extension → |
| FrootAI: Initialize Hooks | Copy guardrails.json (preToolUse policy gates) to your project | ✅ Shipped | Extension → |
| FrootAI: Initialize Prompts | Copy 4 slash commands (/deploy, /test, /review, /evaluate) | ✅ Shipped | Extension → |
| FrootAI: Look Up AI Term | 200+ terms — inline popup with rich definition from curated glossary | ✅ Shipped | Extension → |
| FrootAI: Search Knowledge Base | Full-text search across 18 bundled FROOT modules | ✅ Shipped | Extension → |
| FrootAI: Open Solution Play | View any play in rich webview panel (standalone, offline) | ✅ Shipped | Extension → |
| FrootAI: Show Architecture Pattern | 7 decision guides: RAG, agents, hosting, cost, deterministic AI | ✅ Shipped | Extension → |
| FrootAI: Open Setup Guide | Opens the setup guide on the website | ✅ Shipped | Extension → |
| FrootAI: Browse Solution Plays | Opens the solution plays page on the website | ✅ Shipped | Extension → |
| FrootAI: Open User Guide | Opens the user guide documentation | ✅ Shipped | Extension → |
664KB of curated AI/ML knowledge organized by the FROOT taxonomy — Foundations, Reasoning, Orchestration, Operations, Transformation. Each module is a complete reference accessible via MCP, VS Code, or the website.
| ID | Module Name | Layer | Key Topics | Status | Link |
|---|---|---|---|---|---|
| F1 | GenAI Foundations | 🌱 Foundations | Transformers, attention, tokenization, inference, parameters, context windows | ✅ Shipped | Read → |
| F2 | LLM Landscape & Model Selection | 🌱 Foundations | GPT, Claude, Llama, Gemini, Phi — benchmarks, open vs proprietary | ✅ Shipped | Read → |
| F3 | AI Glossary A–Z | 🌱 Foundations | 200+ AI/ML terms defined — ablation to zero-shot | ✅ Shipped | Read → |
| F4 | .github Agentic OS — 7 Primitives | 🌱 Foundations | Instructions, prompts, agents, skills, hooks, workflows, plugins | ✅ Shipped | Read → |
| R1 | Prompt Engineering & Grounding | 🪵 Reasoning | System messages, few-shot, chain-of-thought, structured output | ✅ Shipped | Read → |
| R2 | RAG Architecture & Retrieval | 🪵 Reasoning | Chunking, embeddings, vector search, hybrid search, reranking | ✅ Shipped | Read → |
| R3 | Making AI Deterministic & Reliable | 🪵 Reasoning | Hallucination reduction, grounding, temperature tuning, guardrails | ✅ Shipped | Read → |
| O1 | Semantic Kernel & Orchestration | 🌿 Orchestration | Plugins, planners, memory, connectors, SK vs LangChain | ✅ Shipped | Read → |
| O2 | AI Agents & Agent Framework | 🌿 Orchestration | Planning, memory, tool use, AutoGen, multi-agent patterns | ✅ Shipped | Read → |
| O3 | MCP, Tools & Function Calling | 🌿 Orchestration | Model Context Protocol, tool schemas, A2A, MCP servers | ✅ Shipped | Read → |
| O4 | Azure AI Platform & Landing Zones | 🍃 Operations | AI Foundry, Model Catalog, deployments, endpoints, enterprise patterns | ✅ Shipped | Read → |
| O5 | AI Infrastructure & Hosting | 🍃 Operations | GPU compute, Container Apps, AKS, App Service, scaling, cost | ✅ Shipped | Read → |
| O6 | Copilot Ecosystem & Low-Code AI | 🍃 Operations | M365 Copilot, Copilot Studio, Power Platform AI, extensibility | ✅ Shipped | Read → |
| T1 | Fine-Tuning & Model Customization | 🍎 Transformation | LoRA, QLoRA, RLHF, DPO, evaluation, MLOps lifecycle | ✅ Shipped | Read → |
| T2 | Responsible AI & Safety | 🍎 Transformation | Content safety, red teaming, guardrails, evaluation frameworks | ✅ Shipped | Read → |
| T3 | Production Architecture Patterns | 🍎 Transformation | Multi-agent hosting, API gateway, latency, cost control, monitoring | ✅ Shipped | Read → |
| QR | Quick Reference Cards | 📎 Reference | Cheat sheets, decision trees, comparison matrices, one-pagers | ✅ Shipped | Read → |
| QZ | Quiz & Assessment | 📎 Reference | Knowledge checks, certification prep, self-assessment questions | ✅ Shipped | Read → |
Two intelligent interfaces for exploring FrootAI. The AI Assistant is a chatbot that recommends plays, compares models, and estimates costs. The Solution Configurator is a 3-question wizard that matches you to the right play.
| Feature | Description | Status | Link |
|---|---|---|---|
| AI Chatbot | Conversational interface — ask about plays, models, costs, setup | 🔄 Preview | Try It → |
| Play Recommendation | Keyword-based play matching (RAG, agent, document, voice, cost) | 🔄 Preview | Try It → |
| Cost Estimation | Scenario-based pricing for dev/staging/production environments | 🔄 Preview | Try It → |
| Model Comparison | GPT-4o vs GPT-4o-mini vs Claude — side-by-side analysis | 🔄 Preview | Try It → |
| MCP Server Guidance | Setup instructions, configuration, troubleshooting in chat | 🔄 Preview | Try It → |
| Solution Configurator | 3-step wizard: What are you building? → Team role? → Complexity? → Play recommendation | ✅ Shipped | Configure → |
| 8 Use Case Categories | Doc processing, Search/RAG, Agents, Voice, Safety, Infra, Ops, ML | ✅ Shipped | Configure → |
| 4 Team Roles | Platform Eng, App Dev, Data/ML, Security — influences play selection | ✅ Shipped | Configure → |
MCP-powered integrations with enterprise platforms. Each partner connector provides tool schemas that let your AI agent read from and write to external systems — all through the same MCP protocol.
| Feature | Description | Status | Link |
|---|---|---|---|
| ServiceNow (ITSM) | Incident & change management — auto-create tickets, escalate P1s, sync resolution notes | 🔄 Preview | Partners → |
| Salesforce (CRM) | Account lookup, case creation/routing, opportunity insights, knowledge search | 🔄 Preview | Partners → |
| SAP (ERP) | Purchase order lookup, invoice processing, material master data, workflow triggers | 🔄 Preview | Partners → |
| Datadog (Monitoring) | Metric queries, APM trace lookup, alert status, dashboard snapshots | 🔄 Preview | Partners → |
| PagerDuty (Incident) | On-call schedule, incident creation, escalation triggers, status page updates | 🔄 Preview | Partners → |
| Jira (Project) | Issue create/update, sprint board queries, backlog grooming, velocity reports | 🔄 Preview | Partners → |
A decentralized marketplace where anyone can publish agents, skills, and prompts. Each plugin is defined by a plugin.json manifest. The 20 built-in solution plays are the first published plugins.
| Feature | Description | Status | Link |
|---|---|---|---|
| plugin.json Manifest | Standard schema: name, version, type, tags, entry, mcp_tools, config, evaluation | ✅ Shipped | Marketplace → |
| 20 Built-in Plugins | All solution plays published as installable plugins with full DevKit + TuneKit | ✅ Shipped | Marketplace → |
| Community Publishing | Register your GitHub repo → plugin.json discovered automatically | 🔄 Preview | Marketplace → |
| Plugin Types | Agents, skills, prompts, workflows, config packs — composable LEGO blocks | 🔄 Preview | Marketplace → |
| Category Index | RAG, ITSM, security, code-review, compliance, analytics, HR, legal, marketing, infra | 🔄 Preview | Marketplace → |
| One-Click Install | VS Code Extension will auto-install plugin files into your workspace | 🔜 Coming Soon | Marketplace → |
Every solution play includes real Azure Bicep templates for provisioning infrastructure. Built on Azure Verified Modules (AVM) patterns — Managed Identity, private endpoints, proper RBAC, no API keys.
| Feature | Description | Status | Link |
|---|---|---|---|
| infra/main.bicep | Main deployment template per play — AI Foundry, AI Search, Container Apps, etc. | ✅ Shipped | Setup → |
| Managed Identity | All services use system-assigned managed identity — zero API keys | ✅ Shipped | Admin Guide → |
| Private Endpoints | VNet integration with private endpoints for AI Search, OpenAI, Storage | ✅ Shipped | Admin Guide → |
| RBAC Assignments | Least-privilege role assignments — Cognitive Services User, Search Index Data Reader | ✅ Shipped | Admin Guide → |
| Key Vault Integration | Secrets stored in Key Vault with managed identity access | ✅ Shipped | Admin Guide → |
| Multi-Region Support | Play 11 (Advanced Landing Zone) supports multi-region with policy governance | ✅ Shipped | Plays → |
| Azure Verified Modules | Following AVM patterns for consistent, well-architected infrastructure | ✅ Shipped | Architecture → |
| GPU Quota Management | Landing zone includes GPU quota requests and capacity planning | ✅ Shipped | Plays → |
Every solution play includes an evaluation pipeline for measuring AI quality before shipping. Test datasets, scoring metrics, and quality gates ensure production readiness.
| Feature | Description | Status | Link |
|---|---|---|---|
| evaluation/eval.py | Main evaluation script — runs test suite, scores responses, generates report | ✅ Shipped | API Ref → |
| Test Datasets | Curated question/answer pairs per play — ground truth for quality measurement | ✅ Shipped | API Ref → |
| Groundedness Score | Measures how well responses are grounded in source documents (target: >0.95) | ✅ Shipped | User Guide → |
| Relevance Score | Measures response relevance to the query (target: >0.90) | ✅ Shipped | User Guide → |
| Coherence Score | Measures logical consistency and readability (target: >0.85) | ✅ Shipped | User Guide → |
| Fluency Score | Measures language quality and naturalness (target: >0.90) | ✅ Shipped | User Guide → |
| Consistency Check | Deterministic agent eval — same input produces same output (target: >95%) | ✅ Shipped | User Guide → |
| Quality Gates | CI/CD integration — block deployment if scores drop below thresholds | ✅ Shipped | Architecture → |
GitHub Actions workflows for automated testing, evaluation, and deployment. The agentic CI/CD layer uses AI to review code, run evals, and enforce guardrails before merging.
| Feature | Description | Status | Link |
|---|---|---|---|
| GitHub Actions CI | Automated lint, build, test on every PR — standard quality gates | ✅ Shipped | Actions → |
| Evaluation on PR | Run eval.py on changed plays — block merge if quality drops | ✅ Shipped | Contributing → |
| Agentic Workflows | Layer 4: AI-driven build → review → tune pipeline in .github/workflows/ | ✅ Shipped | Architecture → |
| Bicep Validation | az deployment validate on infrastructure changes — catch errors early | ✅ Shipped | Admin Guide → |
| npm Publish | Automated npm publish for frootai-mcp on version bump | ✅ Shipped | npm → |
| Extension Publish | Automated VS Code Marketplace publish via vsce | ✅ Shipped | Marketplace → |
| Website Deploy | Docusaurus build + deploy on main branch push | ✅ Shipped | Website → |
The central developer portal for building with FrootAI — guides, API reference, changelogs, and architecture documentation. A one-stop shop for contributors and power users.
| Feature | Description | Status | Link |
|---|---|---|---|
| Admin Guide | Platform setup, MCP configuration, security hardening, troubleshooting | ✅ Shipped | Read → |
| User Guide | End-to-end usage — 13 commands, 16 tools, DevKit/TuneKit deep dive, FAQ | ✅ Shipped | Read → |
| Contributor Guide | Dev setup, repo structure, PR process, code style, testing requirements | ✅ Shipped | Read → |
| API Reference | All 16 MCP tools with schemas, 13 VS Code commands, config file schemas | ✅ Shipped | Read → |
| Architecture Overview | System diagram, 6 architectural layers, data flow, deployment topology | ✅ Shipped | Read → |
| Changelog | Full release history — v2.2.0, v0.9.2, v0.9.0, v2.1.1, v1.0.0 | ✅ Shipped | Read → |
| Quick Start | 3-step getting started: Install Extension → Init DevKit → Deploy | ✅ Shipped | Dev Hub → |
FrootAI is fully open-source under the MIT license. Community contributions are welcome — from new solution plays to knowledge modules to MCP tools.
| Feature | Description | Status | Link |
|---|---|---|---|
| MIT License | Fully permissive — use, modify, distribute freely | ✅ Shipped | License → |
| GitHub Issues | Bug reports, feature requests, and discussion | ✅ Shipped | Issues → |
| PR Template | Structured pull request template for consistent contributions | ✅ Shipped | Template → |
| CONTRIBUTING.md | Contribution guide — setup, conventions, review process | ✅ Shipped | Read → |
| Community Plays | Anyone can submit new solution plays via PR | ✅ Shipped | Contributing → |
| Knowledge Contributions | Add new FROOT modules or improve existing ones | ✅ Shipped | Contributing → |
| Plugin Submissions | Publish community plugins to the marketplace via plugin.json | 🔄 Preview | Marketplace → |
What's next for FrootAI. The roadmap is community-driven — upvote features on GitHub Issues or submit your own ideas.
| Feature | Description | Status | Link |
|---|---|---|---|
| Full AI Chatbot | OpenAI-powered assistant replacing preview keyword matching with real LLM responses | 🔜 Coming Soon | Preview → |
| One-Click Plugin Install | Install community plugins directly from Marketplace into your workspace | 🔜 Coming Soon | Marketplace → |
| Agent Evaluation Dashboard | Visual dashboard for eval scores, trends, and quality gate status | 🔜 Coming Soon | Architecture → |
| Multi-Language MCP | Python and Go MCP server implementations alongside Node.js | 🔜 Coming Soon | MCP → |
| Foundry Agent Hosting | One-click deploy solution plays as hosted Foundry agents | 🔜 Coming Soon | Setup → |
| Prompt A/B Testing | Built-in A/B testing framework for prompt variants across environments | 🔜 Coming Soon | Plays → |
| Cost Dashboard | Real-time Azure cost tracking per solution play deployment | 🔜 Coming Soon | MCP → |
| Enterprise SSO | Entra ID integration for team-based access control | 🔜 Coming Soon | Enterprise → |
| Telemetry & Analytics | App Insights integration for MCP usage analytics and play adoption metrics | 🔜 Coming Soon | Admin Guide → |
| Solution Play Templates | GitHub template repos for instant project scaffolding | 🔜 Coming Soon | Plays → |