World Bank Data360 MCP Server

Instructions

Implementation Reference

• src/world_bank_mcp/server.py:580-615 (handler)

  The MCP tool handler for 'search_local_indicators'. Includes @server.tool() registration decorator, type-hinted parameters defining the input schema, comprehensive docstring, and delegates to the search_local_metadata helper for implementation.
  @server.tool() def search_local_indicators(query: str, limit: int = 20) -> dict[str, Any]: """Search through local metadata for World Bank indicators (instant, offline). This is a FAST alternative to search_datasets that searches through locally cached metadata. Use this to discover what types of data are available before using the API search. How it works: - Searches indicator names, codes, and descriptions - Instant results (no API call) - Returns relevance-ranked matches Search tips: - Use simple keywords: "unemployment", "poverty", "co2", "water" - Search works on indicator names AND descriptions - Case-insensitive - More specific queries = better results Example queries: - "unemployment" → finds all unemployment-related indicators - "mortality infant" → finds infant mortality indicators - "internet" → finds internet usage indicators - "renewable energy" → finds renewable energy indicators Parameters: - query: Search term (e.g., "unemployment", "gdp growth", "water access") - limit: Maximum number of results to return (default: 20) Returns: List of matching indicators with codes, names, descriptions, and relevance scores. Note: This returns indicator codes but NOT database IDs. After finding an indicator, use search_datasets with the indicator name to get the database ID needed for data retrieval. """ return search_local_metadata(query, limit)
• src/world_bank_mcp/server.py:293-354 (helper)

  Core helper function implementing the search logic: loads local metadata, performs fuzzy search on indicator name/code/description with relevance scoring, sorts and limits results, formats output.
  def search_local_metadata(query: str, limit: int = 20) -> dict[str, Any]: """Search through local metadata (fast, offline)""" indicators = load_metadata() if not indicators: return { "success": False, "error": "Metadata file not found. Please ensure metadata_indicators.json exists." } query_lower = query.lower() results = [] # Search through indicators for indicator in indicators: name_lower = indicator["name"].lower() desc_lower = indicator["description"].lower() code_lower = indicator["code"].lower() # Calculate relevance score score = 0 # Exact match in code (highest priority) if query_lower == code_lower: score = 100 # Contains in code elif query_lower in code_lower: score = 90 # Exact word match in name elif query_lower in name_lower.split(): score = 80 # Contains in name (high priority) elif query_lower in name_lower: score = 70 # Contains in description (lower priority) elif query_lower in desc_lower: score = 40 else: continue results.append({ "indicator": indicator["code"], "name": indicator["name"], "description": indicator["description"][:200] + "..." if len(indicator["description"]) > 200 else indicator["description"], "source": indicator["source"][:100] + "..." if len(indicator["source"]) > 100 else indicator["source"], "relevance_score": score }) # Sort by relevance score results.sort(key=lambda x: x["relevance_score"], reverse=True) # Limit results results = results[:limit] return { "success": True, "query": query, "total_matches": len(results), "results": results, "note": "Local search - instant results from cached metadata" }