Enterprise AI Toolkit MCP

An enterprise MCP server that exposes reusable AI, RAG, document intelligence, knowledge graph, SQL, evaluation, and AI architecture capabilities as standardized tools for AI agents.

---

🌟 Vision & Overview

The Enterprise AI Toolkit MCP acts as an enterprise capability gateway between AI agents (Claude, Gemini, OpenAI, custom agents) and backend enterprise AI infrastructure. Instead of coupling an agent to monolithic custom code or fragile API calls, agents can discover and compose standardized tools for document processing, RAG research, knowledge graph queries, natural language SQL, model evaluation, and cloud architecture design.

                         ┌───────────────────────┐
                         │      AI Agent         │
                         │ Claude / Gemini /     │
                         │ OpenAI / Custom Agent │
                         └───────────┬───────────┘
                                     │
                                     │ MCP JSON-RPC
                                     ▼
                    ┌──────────────────────────────┐
                    │   Enterprise AI Toolkit MCP  │
                    │                              │
                    │  MCP Server + Tool Registry  │
                    └──────────────┬───────────────┘
                                   │
          ┌────────────────────────┼────────────────────────┐
          │                        │                        │
          ▼                        ▼                        ▼
 ┌─────────────────┐      ┌─────────────────┐      ┌─────────────────┐
 │ Document Tools  │      │ Knowledge Tools │      │ Data Tools      │
 ├─────────────────┤      ├─────────────────┤      ├─────────────────┤
 │ OCR & Extract   │      │ Vector Search   │      │ SQL Generation  │
 │ Summarization   │      │ Grounded RAG    │      │ SQL Validation  │
 │ Contract Audit  │      │ Knowledge Graph │      │ Execute Query   │
 └─────────────────┘      └─────────────────┘      └─────────────────┘

          ┌────────────────────────┼────────────────────────┐
          │                        │                        │
          ▼                        ▼                        ▼
 ┌─────────────────┐      ┌─────────────────┐      ┌─────────────────┐
 │ LLM Tools       │      │ Evaluation      │      │ Architecture   │
 ├─────────────────┤      ├─────────────────┤      ├─────────────────┤
 │ LLM Evaluation  │      │ Faithfulness    │      │ AI Architecture│
 │ Cost Estimator  │      │ Groundedness    │      │ TCO Cost Calc   │
 │ Prompt Guard    │      │ Latency & Cost  │      │ Tech Stack      │
 └─────────────────┘      └─────────────────┘      └─────────────────┘

---

🚀 Key Features

• 16 Standardized MCP Tools: Complete coverage for Document Intelligence, RAG, Knowledge Graph, SQL Intelligence, LLM Evaluation, Cost Estimation, and Cloud Architecture.
• 100% Demo Mode / Zero API Key Requirement: Runs out of the box using intelligent offline MockLLMProvider, local vector index, in-memory Knowledge Graph, and SQLite database.
• LLM Provider Agnostic: Easily toggle between Google Gemini (gemini-2.5-flash), OpenAI (gpt-4o-mini), and local mock provider.
• Enterprise Security & Governance:
  • PromptGuard: Prompt injection threat detection and token efficiency analysis.
  • SQLGuard: Enforces read-only execution, blocks multi-statements, and catches dangerous keywords (DROP, DELETE, TRUNCATE).
  • InputValidator: Path traversal protection and upload size guards.
• MCP Resources & Prompts: Exposes resource://enterprise-ai/catalog, configuration, model-registry, tool-metrics, and standardized prompts (enterprise_rag_analysis, enterprise_sql_analysis, enterprise_contract_review, enterprise_ai_architecture).

---

🛠️ MCP Tools Reference

Category	Tool Name	Description	Risk Level
Documents	extract_document	Extracts text, pages, metadata, headings, and chunks	LOW
Documents	summarize_document	Generates executive summaries, key takeaways, and action items	LOW
Documents	analyze_contract	Audits legal contracts for terms, obligations, risks, and risk score	MEDIUM
Knowledge	search_knowledge	Semantic vector search over enterprise document chunks	LOW
Knowledge	generate_rag_answer	Grounded RAG research with source citations	LOW
Knowledge	query_knowledge_graph	Queries and traverses entity-relationship Knowledge Graph	LOW
Data	generate_sql	Translates natural language into safe SQL queries	MEDIUM
Data	validate_sql	Validates SQL syntax against security policies and injection	LOW
Data	execute_sql	Safely executes read-only SQL queries against database	MEDIUM
Data	analyze_database_schema	Inspects database schema, foreign keys, and optimization tips	LOW
LLM	evaluate_llm	Evaluates LLM responses across correctness, relevance, faithfulness	LOW
LLM	estimate_llm_cost	Estimates daily, monthly, and annual LLM pricing	LOW
LLM	analyze_prompt	Analyzes prompt for security risks and injection threats	LOW
LLM	recommend_llm	Recommends optimal LLM based on budget, latency, and privacy	LOW
Architecture	design_ai_architecture	Generates enterprise AI architecture design with Mermaid diagram	LOW
Architecture	estimate_ai_solution_cost	Calculates TCO for enterprise GenAI solution	LOW

---

💻 Quick Start & Installation

1. Install local package

pip install -e .

2. Run CLI commands

# List all 16 registered tools
enterprise-ai-mcp list-tools

# Check server health
enterprise-ai-mcp health

# Run E2E BFSI Demo Workflow
enterprise-ai-mcp demo

3. Run MCP Server in stdio mode

enterprise-ai-mcp server --transport stdio

---

🧪 Testing

The repository includes a unit and integration test suite runnable without external API keys:

python -m pytest

---

🐳 Docker Deployment

# Docker Compose
docker-compose up -d

# Docker CLI
docker build -t enterprise-ai-mcp .
docker run -p 8000:8000 enterprise-ai-mcp

---

📝 License

Distributed under the MIT License.