The server exposes 10 data center intelligence tools with structured schemas and clear descriptions. Most tools follow verb_noun naming convention (search_facilities, get_facility, list_transactions, get_market_intel, get_news, analyze_site, dchub_market_intel, dchub_search_facilities, dchub_grid, dchub_market). All tools are READ_ONLY and have documented input parameters with types and descriptions. However, output schemas are not explicitly documented, only input schemas are visible in the provided code. Parameter descriptions are adequate but could be more detailed regarding constraints and use cases. Tool descriptions average ~150-180 chars, which is within the 10-1024 baseline but some lack depth about when to use them vs related tools. Error handling details are absent from all tools. No evidence of LLM-specific guidance like 'call this discovery tool first' or dependency hints.

• Add comprehensive output schema documentation for every tool. Specify field names, types (string, integer, object, array), and what each field represents. Example: 'Returns an object with fields: {facilities: [{ id, name, provider, city, country, power_mw, cooling, fiber_count }], total_count: integer, citation_url: string}'.
• Consolidate duplicate tools. Choose one canonical name per function (e.g., 'search_facilities' or 'dchub_search_facilities', not both). Clearly document when tools serve the same purpose but different formats (REST vs MCP integrations).
• Add error handling guidance to each tool description. E.g., 'Returns 404 if no facilities match. Try broadening your search filters. Returns 429 if rate-limited; retry after 30 seconds.'
• Document parameter constraints explicitly. For 'limit', add 'max 100' or 'max 50'. For 'radius_miles', add 'range 1-100, default 25'. For 'query', add 'case-insensitive, supports wildcards'.
• Add discovery hints to related tools. E.g., in 'get_facility', add 'Use search_facilities() first to find facility IDs if you only have a name or location.'
• Provide pagination details. State whether 'limit' returns exactly N results, up to N results, or uses offset/cursor. Include 'total_count' and 'next_cursor' or 'has_more' fields in output schema.
• Unify tool definitions across frameworks. Maintain a single canonical tool manifest (e.g., tools.json) that is shared by the MCP SDK, Langchain, Cohere, and LlamaIndex integrations. Use generated SDKs to avoid divergence.

Inferred effective spec: 2026-07-28+.

• Relies on Stateful initialize / Mcp-Session-Id (removed; protocol is stateless) - make each request self-contained