WonderSearch

Teams and companies accumulate documents, research, and internal knowledge that is hard to search because people do not know the exact phrasing used in the source material. Traditional semantic search setups require building embedding pipelines and ingesting data upfront, which adds cost and complexity before anyone can even run a search. Customers hunting for a policy or product detail in a help center often cannot find the right answer without manual digging. AI assistants also need a way to pull in relevant context from a company's own files rather than relying on general knowledge.

WonderSearch is described as a serverless sparse retrieval solution for semantic search that lets users upload files (PDF, TXT, MD) and start asking questions in natural language without creating embeddings or paying upfront ingestion fees. It returns relevant passages with source references, accessible through a REST API, Python SDK, CLI, or MCP, so it can be used directly, embedded into a product's search or help center, or used to feed context to AI assistants like those from OpenAI or Anthropic. Pricing is per query rather than for idle compute, with storage billed separately at object storage rates. The product also offers a configurable 'search effort' setting (small, medium, large) so users can trade off speed versus thoroughness depending on the complexity of the question.