MCP Servers

Overview#

All 13 Python products expose Model Context Protocol (MCP) servers via FastApiMCP. This enables AI assistants, development tools, and other MCP clients to interact with Anar products programmatically -- querying data, triggering actions, and receiving structured responses.

What is MCP?#

The Model Context Protocol is an open standard for connecting AI models to external tools and data sources. Each Anar product registers its API endpoints as MCP tools, allowing AI assistants to discover and invoke them without custom integration code.

MCP Endpoints#

Every Python product mounts its MCP server at the /mcp path:

Setting Up MCP in a Product#

FastApiMCP automatically exposes all FastAPI routes as MCP tools. The setup in each product's main.py:

from fastapi import FastAPI
from fastapi_mcp import FastApiMCP

app = FastAPI(title="Anar Guard")

# ... define routes ...

mcp = FastApiMCP(app, name="anar-guard", description="AI Governance & Compliance")
mcp.mount()

The mount() call adds the /mcp endpoint to the FastAPI app. All routes registered before mount() are automatically exposed as MCP tools.

Connecting from Claude Desktop#

Add the MCP server to your Claude Desktop configuration file (claude_desktop_config.json):

{
  "mcpServers": {
    "anar-guard": {
      "url": "http://localhost:8002/mcp"
    },
    "anar-chat": {
      "url": "http://localhost:8001/mcp"
    },
    "anar-translate": {
      "url": "http://localhost:8012/mcp"
    }
  }
}

After restarting Claude Desktop, the tools from each Anar product appear in the tool list. You can then ask Claude to scan content with Guard, query knowledge bases in Chat, or translate documents.

Connecting from Claude Code#

In Claude Code, add Anar MCP servers to your project configuration:

claude mcp add anar-guard http://localhost:8002/mcp
claude mcp add anar-chat http://localhost:8001/mcp
claude mcp add anar-docs http://localhost:8008/mcp

Available Tools per Product#

Each product exposes its API routes as MCP tools. Here are key tools from each product:

Guard#

• scan_content -- Run text through policy engine and safety scanning
• detect_pii -- Identify PII entities (national IDs, phone numbers, IBAN)
• list_policies -- List configured governance policies
• get_compliance_report -- Generate GCC compliance report

Chat#

• query_knowledge_base -- RAG query against a knowledge base
• list_conversations -- List chat conversations
• create_conversation -- Start a new conversation
• ingest_document -- Upload a document to the knowledge base

Voice#

• transcribe_audio -- Transcribe audio with dialect detection
• list_transcriptions -- List past transcriptions

Translate#

• translate_text -- Translate text between languages
• list_glossaries -- List translation glossaries
• import_tmx -- Import translation memory from TMX

Docs#

• process_document -- OCR and extract text from a document
• classify_document -- Classify document type
• list_documents -- List processed documents

Flow#

• list_workflows -- List available workflows
• trigger_workflow -- Execute a workflow with inputs
• get_execution_status -- Check workflow execution status

Comply#

• run_assessment -- Run compliance assessment against a framework
• list_frameworks -- List available compliance frameworks

Present#

• generate_presentation -- Generate slides from a prompt
• export_pptx -- Export presentation as PPTX
• list_presentations -- List generated presentations

Minutes#

• create_meeting -- Create a meeting record
• upload_audio -- Upload meeting audio for transcription
• get_summary -- Get meeting summary with action items

Procure#

• analyze_tender -- Analyze a tender document
• score_bids -- Score and rank bids
• list_tenders -- List tenders

Authentication#

MCP requests follow the same JWT authentication as regular API calls. Include the Bearer token in your MCP client configuration if the product requires authentication:

{
  "mcpServers": {
    "anar-guard": {
      "url": "http://localhost:8002/mcp",
      "headers": {
        "Authorization": "Bearer eyJhbGciOiJIUzI1NiIs..."
      }
    }
  }
}

Combining Multiple Products#

The real power of MCP comes from connecting multiple Anar products to a single AI assistant. For example, a Claude assistant with access to Guard, Chat, and Docs can:

1. Accept a document upload via Docs
2. Extract text and tables with OCR
3. Scan the content for PII with Guard
4. Ingest the clean content into Chat's knowledge base

All orchestrated through natural language instructions to the assistant.