Structured entity data and the questions people actually ask around a topic give a model something to reason with, not just pages to summarise — that is the difference from a plain results list. Pick the research depth deliberately: the deep setting is described as pulling ten or more sources over fifteen to twenty minutes, which is a different kind of request from a lookup.
A TypeScript MCP server built on the Serper API. It offers a plain search with rich structured results, and a deep research mode that breaks a question into sub-queries, runs several searches, and synthesises the findings with citations.
- Search returning knowledge graph entries, organic results with position and snippet, People Also Ask entries, and related searches
- Search parameters for result count, country and language targeting, and autocorrect
- Deep research that generates its own sub-queries, gathers from multiple sources, and formats the answer with references
- Three research depths: basic for a quick overview, standard for a balanced pass, deep for exhaustive work, with a cap on sources you set
- Rate limiting and caching, and full TypeScript type definitions
A Serper API key set as `SERPER_API_KEY`. Clone, install with pnpm and build, then point your client at the built entry point. Quality metrics are collected by default and can be redirected or disabled through their own environment variables.