FrootAI API Reference

> Technical reference for all public APIs — MCP tools, VS Code commands, config schemas, and Agentic OS files.

1. MCP Server Tools (16 tools)

The FrootAI MCP Server exposes tools over the Model Context Protocol (stdio transport). All tools return JSON-RPC 2.0 responses.

1.1 Static Tools

get_module

Retrieve a full knowledge module by ID or name.

Output: { content: [{ type: "text", text: "<full module content>" }] }

Example:

{
  "name": "get_module",
  "arguments": { "module": "RAG-Architecture" }
}

list_modules

List all available knowledge modules with metadata.

Output: { content: [{ type: "text", text: "<JSON array of modules>" }] }

Each module entry:

{
  "id": "F1",
  "name": "GenAI-Foundations",
  "layer": "Foundations",
  "title": "GenAI Foundations",
  "sizeBytes": 24500
}

search_knowledge

Full-text search across all knowledge modules.

Output: { content: [{ type: "text", text: "<JSON array of matches>" }] }

Each match:

{
  "module": "RAG-Architecture",
  "section": "## Vector Databases",
  "snippet": "...matching text with context...",
  "score": 0.87
}

lookup_term

Look up an AI term in the glossary (comprehensive glossary).

Output: { content: [{ type: "text", text: "<definition and context>" }] }

1.2 Live Tools

fetch_azure_docs

Retrieve current Azure documentation for a service or topic.

Output: { content: [{ type: "text", text: "<documentation content>" }] }

> Note: Requires network connectivity.

fetch_external_mcp

Query external MCP server registries and catalogs.

Output: { content: [{ type: "text", text: "<JSON array of MCP servers>" }] }

1.3 Chain Tools

agent_build

Generate a new agent scaffold based on a scenario description.

Output: { content: [{ type: "text", text: "<generated agent.md + config>" }] }

agent_review

Review an agent’s configuration for quality, security, and completeness.

Output: { content: [{ type: "text", text: "<review findings and recommendations>" }] }

agent_tune

Optimize agent parameters based on evaluation results.

Output: { content: [{ type: "text", text: "<optimized config + explanation>" }] }

1.4 AI Ecosystem Tools

get_architecture_pattern

Get architecture patterns for a given scenario.

Output: { content: [{ type: "text", text: "<pattern description with diagram>" }] }

get_froot_overview

Get an overview of the FrootAI platform and its components.

Output: { content: [{ type: "text", text: "<platform overview>" }] }

get_github_agentic_os

Explain the .github Agentic OS — files, primitives, and composition.

Output: { content: [{ type: "text", text: "<Agentic OS documentation>" }] }

list_community_plays

Browse community-contributed solution plays.

Output: { content: [{ type: "text", text: "<JSON array of plays>" }] }

get_ai_model_guidance

Get model selection guidance and comparison data.

Output: { content: [{ type: "text", text: "<model comparison and recommendation>" }] }

2. VS Code Extension Commands (13 commands)

All commands are prefixed with frootai. and appear in the Command Palette as FROOT:.

3. Config File Schemas

3.1 openai.json

{
  "_comment": "Azure OpenAI deployment configuration",
  "deployment_name": "gpt-4o",
  "api_version": "2024-08-01-preview",
  "temperature": 0.7,
  "max_tokens": 4096,
  "top_p": 0.95,
  "frequency_penalty": 0,
  "presence_penalty": 0
}

3.2 guardrails.json

{
  "_comment": "Content safety and guardrail configuration",
  "max_input_tokens": 8192,
  "max_output_tokens": 4096,
  "blocked_topics": ["violence", "self-harm", "illegal-activity"],
  "content_filter_level": "medium",
  "rate_limit_rpm": 60,
  "require_grounding": true,
  "citation_required": true
}

3.3 agents.json

{
  "_comment": "Agent routing and model configuration",
  "default_model": "gpt-4o",
  "fallback_model": "gpt-4o-mini",
  "routing": {
    "complex_reasoning": "gpt-4o",
    "simple_qa": "gpt-4o-mini",
    "code_generation": "gpt-4o"
  },
  "temperature_overrides": {
    "code_generation": 0.2,
    "creative_writing": 0.9
  },
  "max_retries": 3,
  "timeout_ms": 30000
}

3.4 model-comparison.json

{
  "_comment": "Model comparison data for selection guidance",
  "models": [
    {
      "name": "gpt-4o",
      "provider": "Azure OpenAI",
      "context_window": 128000,
      "cost_per_1k_input": 0.005,
      "cost_per_1k_output": 0.015,
      "latency_p50_ms": 800,
      "strengths": ["reasoning", "code", "multimodal"]
    }
  ]
}

3.5 search.json

{
  "_comment": "Search configuration for RAG pipelines",
  "method": "hybrid",
  "top_k": 5,
  "min_score": 0.7,
  "reranker": true,
  "embedding_model": "text-embedding-3-large",
  "dimensions": 3072
}

3.6 chunking.json

{
  "_comment": "Document chunking strategy",
  "strategy": "semantic",
  "chunk_size_tokens": 512,
  "overlap_tokens": 64,
  "separators": ["\n## ", "\n### ", "\n\n", "\n"],
  "preserve_markdown_structure": true
}

4. Plugin Manifest (plugin.json)

The root plugin.json describes the FrootAI plugin for marketplace and registry listing.

{
  "name": "frootai",
  "version": "2.2.0",
  "description": "BIY AI Kit — From the Roots to the Fruits",
  "author": "FrootAI Contributors",
  "license": "MIT",
  "repository": "https://github.com/frootai/frootai",
  "components": {
    "mcp-server": { "version": "2.2.0", "tools": 16 },
    "vscode-extension": { "version": "0.9.2", "commands": 13 },
    "website": { "pages": 13 },
    "knowledge-modules": { "count": 18 },
    "solution-plays": { "count": 20 }
  },
  "keywords": ["ai", "mcp", "agents", "azure", "rag", "froot"],
  "categories": ["AI", "Developer Tools", "Azure"]
}

5. .github Agentic OS File Reference (19 files)

The Agentic OS is a scaffolded set of files that make any project agent-ready. Organized in 4 layers:

Layer 1: Agent Identity

Layer 2: Prompt Library

Layer 3: CI/CD Automation

Layer 4: Collaboration Templates

Composition

These files compose through 7 primitives:

1. Agent Rules → agent.md
2. Context → copilot-instructions.md
3. Prompts → prompts/*.prompt.md
4. Workflows → workflows/*.yml
5. Templates → ISSUE_TEMPLATE/, pull_request_template.md
6. Config → config/*.json
7. Evaluation → evaluation/

Each primitive is independent but they strengthen each other. An agent with rules + context + prompts is significantly more capable than rules alone.

Scripts & Automation

deploy-play.sh / deploy-play.ps1

Deploy any solution play end-to-end (infra + config + eval).

Usage:

./scripts/deploy-play.sh <play-number> [--resource-group <rg>] [--skip-eval]

rebuild-knowledge.sh / rebuild-knowledge.ps1

Rebuild knowledge.json from docs/ and sync to VS Code extension.

Usage:

./scripts/rebuild-knowledge.sh [--publish]

export-skills.sh / export-skills.ps1

Export FROOT modules as .github/skills/ folders for GitHub Copilot.

Usage:

./scripts/export-skills.sh <module-id>   # e.g., F1, R2, O3
./scripts/export-skills.sh --all          # export all knowledge modules

auto-update.js (MCP Server)

Knowledge auto-refresh module. Checks if knowledge.json is older than 7 days and fetches latest from GitHub.

import { getLatestKnowledge } from './auto-update.js';
const knowledge = await getLatestKnowledge();

Azure Integration

azure.yaml

Configuration for azd up deployment. Place in the root of any solution play.

registry-entry.json

MCP ecosystem registry entry for FrootAI. Located at mcp-server/registry-entry.json.

agent-card.json

A2A (Agent-to-Agent) protocol card. Located at mcp-server/agent-card.json.

> Next: Admin Guide · User Guide · Architecture Overview