When an agent is asked “should I buy this?”, the open web gives it marketing pages, affiliate listicles, and review spam. Buyary gives it the synthesis: a composite Buyary Score with expert/user/value sub-scores, verified claims, and per-source consensus — every field traceable to independent sources, with raw reviews and store listings deliberately omitted.

The whole surface is two tools, callable over remote MCP or plain REST: get_product_intelligence for one product, and get_product_intelligence_batch for up to five at once. Issue a key once — it works on both protocols — point your client at the service, and your agent answers buying questions from the Evidence Layer instead of guessing. When nothing confidently matches, it says so rather than returning a wrong product.

Where to go next