MCP Tools Reference

The FrootAI MCP Server exposes 25 tools that any MCP-compatible client (VS Code Copilot, Claude Desktop, Cursor, etc.) can invoke. Install and start using them immediately:


npx frootai-mcp@latest

💡

Quick Start

Add the MCP server to your VS Code mcp.json or Claude Desktop config, then call any tool by name. See the CLI Commands page for alternative CLI usage.

Complete Tool Catalog

Tool	Description	Category
`search_knowledge`	Search across all 17 FROOT modules	Knowledge
`get_module`	Get full content of a specific module (F1–T3)	Knowledge
`lookup_term`	Look up AI/ML term in glossary (200+ terms)	Knowledge
`list_modules`	List all modules by FROOT layer	Knowledge
`semantic_search_plays`	Natural language search for solution plays	Plays
`get_play_detail`	Detailed architecture for a specific play	Plays
`list_community_plays`	List all solution plays with status	Plays
`compare_plays`	Side-by-side play comparison	Plays
`agent_build`	Builder agent — implementation guidance	Agents
`agent_review`	Reviewer agent — security + quality review	Agents
`agent_tune`	Tuner agent — production readiness validation	Agents
`compare_models`	AI model comparison for use case	Models
`get_model_catalog`	List Azure OpenAI models with pricing	Models
`estimate_cost`	Calculate monthly Azure costs for a play	Cost
`get_azure_pricing`	Azure AI pricing for scenarios	Cost
`run_evaluation`	Check AI quality scores against thresholds	Evaluation
`validate_config`	Validate TuneKit config files	Validation
`generate_architecture_diagram`	Mermaid.js architecture diagram	Architecture
`get_architecture_pattern`	Architecture guidance for scenarios	Architecture
`embedding_playground`	Compare text similarity	Learning
`list_primitives`	Browse agents, skills, hooks, plugins, etc.	Primitives
`fetch_azure_docs`	Latest Azure documentation	Azure
`fetch_external_mcp`	Search for external MCP servers	MCP
`get_froot_overview`	Complete FROOT framework overview	Overview
`get_github_agentic_os`	.github Agentic OS guide	Reference

Detailed Tool Reference

1. `search_knowledge`

Search across all 17 FrootAI modules for a topic. Returns relevant sections matching the query.

Parameters:

Parameter	Type	Required	Default	Description
`query`	string	✅	—	Natural language search query
`max_results`	number	❌	5	Maximum matching sections to return

Example call:


{
  "tool": "search_knowledge",
  "arguments": {
    "query": "how to reduce hallucination in RAG",
    "max_results": 3
  }
}

Example response:


{
  "results": [
    {
      "module": "R2 — RAG",
      "section": "Grounding & Citation",
      "content": "Use groundedness checks (score ≥ 4.0)..."
    }
  ]
}

2. `semantic_search_plays`

Describe what you want to build in natural language and get the top matching solution plays ranked by relevance.

Parameters:

Parameter	Type	Required	Default	Description
`query`	string	✅	—	What you want to build
`top_k`	number	❌	3	Number of results (max 5)

Example call:


{
  "tool": "semantic_search_plays",
  "arguments": {
    "query": "process invoices and extract line items",
    "top_k": 3
  }
}

Example response:


{
  "matches": [
    { "play": "06-document-intelligence", "confidence": 0.94 },
    { "play": "15-document-processing", "confidence": 0.87 },
    { "play": "01-enterprise-rag", "confidence": 0.62 }
  ]
}

ℹ️

For full play details after finding a match, chain with get_play_detail passing the play number. See the fai-manifest.json spec for how plays are wired.

3. `agent_build`

Triggers the Builder agent — returns implementation guidelines based on FrootAI best practices, then suggests review. Part of the Build → Review → Tune chain.

Parameters:

Parameter	Type	Required	Description
`task`	string	✅	What the user wants to build

Example call:


{
  "tool": "agent_build",
  "arguments": {
    "task": "IT ticket classification API with Azure OpenAI"
  }
}

⚠️

After building, always follow up with agent_review and then agent_tune to complete the full quality chain. Skipping review or tuning may leave security or production-readiness gaps.

4. `compare_models`

Side-by-side comparison of AI models for a specific use case. Recommends the best model based on your priority.

Parameters:

Parameter	Type	Required	Default	Description
`useCase`	string	✅	—	What you’re building
`priority`	enum	❌	`quality`	`cost`, `quality`, `speed`, or `context`

Example call:


{
  "tool": "compare_models",
  "arguments": {
    "useCase": "RAG chatbot with 50-page documents",
    "priority": "cost"
  }
}

5. `estimate_cost`

Calculate itemized monthly Azure costs for any solution play at dev or production scale.

Parameters:

Parameter	Type	Required	Default	Description
`play`	string	✅	—	Play number: `01`–`20`
`scale`	enum	❌	`dev`	`dev` or `prod`

Example call:


{
  "tool": "estimate_cost",
  "arguments": {
    "play": "01",
    "scale": "prod"
  }
}

Example response:


{
  "play": "01-enterprise-rag",
  "scale": "prod",
  "monthly_total": "$1,240",
  "breakdown": {
    "Azure OpenAI (GPT-4o)": "$800",
    "Azure AI Search (S1)": "$250",
    "Azure App Service (P1v3)": "$140",
    "Azure Cosmos DB (Serverless)": "$50"
  }
}

CLI Commands — command-line alternatives to MCP tools
JSON Schemas — validation schemas for all FrootAI primitives
fai-manifest.json — the wiring spec that ties plays together

MCP Tools Reference

Complete Tool Catalog

Detailed Tool Reference

1. search_knowledge

2. semantic_search_plays

3. agent_build

4. compare_models

5. estimate_cost

Related Pages

1. `search_knowledge`

2. `semantic_search_plays`

3. `agent_build`

4. `compare_models`

5. `estimate_cost`