When a user asks a product question, identify the product and locate its official documentation: manufacturer spec sheet PDF, product listing description, user manual, or a structured product attributes API response.
Fetch and index the documentation: extract text from PDFs, parse structured attribute tables from HTML product pages, and store as a searchable document.
Use semantic search or structured attribute lookup to find the section of the documentation most relevant to the question (e.g., for a battery life question, search for sections containing 'battery', 'mAh', 'hours', 'standby').
Generate an answer grounded in the extracted text; cite the specific document section and page number so the user can verify; do not extrapolate or infer specifications not stated in the source.
If the documentation does not answer the question, surface the original document to the user and suggest contacting the manufacturer directly rather than guessing.
Cache extracted document content indexed by product ID and document version to avoid re-fetching on repeated queries.
Known gotchas
Spec sheets frequently contain model-family-level information that may not apply to a specific variant; always confirm the exact model number on the document matches the product in question.
Manufacturer spec PDFs can be locked or use image-only pages that are not text-extractable; have an OCR fallback for image-based PDFs.
Marketing copy on product pages overstates capabilities; prefer the technical specification table over the feature bullet list when answering precise technical questions.
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp