| Web search | Ships. Web search results via /web. | Ships. Search API returns ranked web results. | Ships. Search API supports neural and keyword retrieval. | Ships. Search API returns results from supported engines. | Ships. Search API returns ranked web results. | Ships. Search endpoint returns web results. |
|---|
| Dedicated social APIs | Ships. Dedicated X, TikTok, Facebook and Instagram endpoints. | Not established in reviewed docs. No dedicated platform endpoints identified in the reviewed Search API docs; web results may still include social pages. | Not established in reviewed docs. No dedicated platform endpoints identified in the reviewed Search API docs; indexed social pages are a different capability. | Partial. Dedicated Instagram and Facebook profile APIs are documented; this is not the same four-platform set. | Not established in reviewed docs. No dedicated platform endpoints identified in the reviewed Search API docs. | Not established in reviewed docs. The reviewed search docs do not establish dedicated social-platform endpoints. |
|---|
| AI answer + source links | Ships. AI Search returns an optional summary and result source links in one response; inline citations are not promised. | Ships. Search can return an optional answer with result URLs in one response; Research also offers formatted citations. | Ships. Answer API returns generated answers and citations. | Partial. Parses Google's AI Overview and references; it does not generate the answer. | Ships. Agent API returns generated answers with source citations. | Partial. Agent can return structured findings with source URLs when the output schema requests them; this is not a standard cited Q&A response. |
|---|
| Extract a supplied URL | Ships. /web/extract accepts a URL and returns extracted text or HTML. | Ships. Extract API accepts URLs and returns page content. | Ships. Contents API retrieves page text from supplied URLs. | Not established in reviewed docs. Search-result parsing does not establish a general endpoint for extracting an arbitrary supplied URL. | Not established in reviewed docs. Search returns content for its results; the reviewed docs do not establish arbitrary URL extraction. | Ships. Scrape API extracts a supplied URL. |
|---|
| Crawl linked site pages | Not established in reviewed docs. /web/crawl is a deprecated alias for single-URL /web/extract; a linked-site crawler is not documented here. | Ships. Crawl API follows links from a starting URL. | Partial. Contents can fetch a bounded set of linked subpages; it does not enumerate an entire site. | Not established in reviewed docs. The reviewed Search API docs do not establish linked-site crawling. | Not established in reviewed docs. The reviewed Search API docs do not establish linked-site crawling. | Ships. Crawl API follows links across a site. |
|---|
| MCP server | Ships. Official MCP integration. | Ships. Remote MCP server. | Ships. Exa MCP server. | Ships. Official MCP integration. | Ships. Official MCP integration. | Ships. Hosted MCP server; the keyless setup exposes Search, Scrape and Parse. |
|---|
| Cost or usage in response | Ships. Successful billable responses expose exact USD cost in the body or X-Desearch-Cost-Usd header. | Partial. Search can include credits used when include_usage is enabled; no per-call USD amount is shown. | Ships. Search and Contents responses expose estimated costDollars; billing uses usage counters. | Not established in reviewed docs. A per-request cost or credit field is not established in the reviewed Search API response docs. | Partial. Agent and Sonar responses show usage.cost.total_cost; this is not established for the raw Search API. | Partial. Agent job status shows creditsUsed; this is not a universal per-request USD cost field. |
|---|