What GlaziersEdge Knowledge MCP server does
The GlaziersEdge Knowledge MCP server connects AI agents to a hosted reference for US commercial glass and glazing. Its coverage includes glass types and terminology, insulating-glass performance, low-E and reflective coatings, safety glazing, energy requirements, impact and fire-rated systems, wind-load topics, storefront and curtain wall systems, butt-glazed assemblies, sealants, testing, specifications, warranties, lead times, and estimating considerations.
The knowledge base contains 32 topics. It also includes 1-inch insulating-glass center-of-glass data from Vitro, Guardian SunGuard, and Viracon; coating equivalents between manufacturers; aluminum framing cross-references; and 2026 US installed budget ranges for selected commercial glazing scopes.
How it works
The service uses Streamable HTTP at https://glazing-mcp.glazingkb.workers.dev/mcp. It is hosted, read-only, and does not require an account, API key, request headers, or local installation. Agents can begin with list_topics, then use get_topic for complete topic content or a specific section. Long topic responses provide a continuation hint for paging.
Free-text questions are handled by search_glazing_knowledge, which uses BM25 relevance ranking and returns snippets tied to topic and section identifiers. For numeric insulating-glass comparisons, use lookup_glass_performance; for substitution screening, use the glass or aluminum cross-reference tools; and for preliminary cost checks, use budget_price_range.
The GlaziersEdge Knowledge MCP server preserves confidence labels in its responses and exposes structured output alongside markdown text. Prompts named glazing_question and or_equal_substitution are available, as are the glaziersedge://topics and glaziersedge://disclaimer resources.
Setup and configuration
Add the endpoint directly to clients that support remote HTTP MCP servers. The repository contains connection information rather than a locally run implementation. Claude Desktop can use the mcp-remote bridge, which requires Node.js; clients with local-only MCP support can use the same bridge command.
Named connection examples are provided for Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, Gemini CLI, and Codex CLI. No environment variables or server-side configuration are required. An optional detail_level=brief URL parameter removes the repeated confidence-tag legend from text responses while retaining tags on individual results.
Tools and capabilities
- Browse the topic index and unverified-item counts with
list_topics.
- Retrieve full topics or sections with tables and confidence labels using
get_topic.
- Search code limits, definitions, standards, estimating issues, and other glazing subjects with
search_glazing_knowledge.
- Filter 1-inch IGU center-of-glass values by manufacturer, product, substrate, SHGC, VLT, U-value, or LSG with
lookup_glass_performance.
- Compare coated-glass products across manufacturers with match ratings using
cross_reference_glass.
- Compare storefront, curtain wall, entrance, window, and specialty framing series with
cross_reference_aluminum_system.
- Request early-stage installed budget ranges by scope and unit with
budget_price_range.
Limitations and notes
Performance results are center-of-glass values, not NFRC whole-product ratings. Equivalents are screening aids, not proof that systems are identical; check glazing plane, depth, glass thickness, thermal tier, structural performance, and project requirements before substitution.
Budget outputs are experienced-estimator [Expert] ranges for early planning and sanity checks, not bid prices. Information marked [UNVERIFIED] or [inference] requires confirmation. The service is reference material rather than engineering, legal, or code-compliance advice. Queries are logged in anonymized, redacted form for service improvement, so confidential project information should not be entered.