Google Search API
Google's programmatic search surface is in transition. The Custom Search JSON API is closed to new customers and retires January 1, 2027; Vertex AI Search in Google Cloud is the supported successor with REST and gRPC clients. There is no official general-purpose SERP API.
Frequently asked questions
Google Search scores A on Supergood's API Report Card. Google's programmatic search surface is in transition. The Custom Search JSON API is closed to new customers and retires January 1, 2027; Vertex AI Search in Google Cloud is the supported successor with REST and gRPC clients. There is no official general-purpose SERP API.
Google Search has a workable official integration path. Most engineering teams can build against it directly. Open API: self-serve, documented, with SDKs
Google Search is the namesake web search product operated by Google LLC (Alphabet), and the dominant general-purpose search engine globally with roughly ~90% worldwide market share.
No single industry vertical, Google Search is horizontal infrastructure. In a B2B/developer context, a team configures a Programmable Search Engine (selecting domains, refinements, image-search options), provisions a Google Cloud API key, and calls the Custom Search JSON API endpoint to retrieve up to 10 results per request (capped at 100 results per query via pagination).
Maximum at the consumer level, Google Search handles on the order of 8.5 billion+ queries per day and is the default search engine for the majority of internet users worldwide.
For B2B SaaS apps that embedded the Custom Search JSON API into knowledge-base search, in-app site search, or AI-app retrieval pipelines, the API is on a critical path: silent failures or quota exhaustion translate directly into empty results, broken search UIs, and degraded RAG quality in production.
Google Search the product launched in September 1998; the Custom Search JSON API and Programmable Search Engine (formerly Google Custom Search Engine, CSE) have been around since the mid-2000s and feel it, the configuration UI, the OpenSearch 1.1 response shape, and the developer console workflows are all visibly legacy compared to modern AI-native search APIs.
Supergood's grades measure one thing: can a customer's engineering team get their own data out programmatically? Supergood checks six things (whether a real API exists, how access is gated, data coverage, auth quality, docs and developer experience, and stability) and rolls them into a letter grade. Supergood re-verifies grades, and they only move on evidence.
Google Search has partial MCP support. Google publishes no official MCP server for Search. Community servers cover it, including the actively maintained Google Workspace MCP with Custom Search tools, but nothing vendor supported exists.
Partially. MCP support for Google Search exists but is incomplete, so what MCP clients like Claude, Cursor, and Codex can reach depends on what the available server covers.