Exa now has two ways into Google's Gemini stack. The search company announced both integrations on its company blog, tying them to last week's Google Cloud Next.
Grounding with Exa Web Search lets Gemini pull fresh results from Exa's API at request time. Per Google's Vertex AI docs, the feature is in Restricted Preview: teams need to ask their Google Cloud account rep for allowlist access. Default rate limit is 200 prompts per minute. Exa says its Highlights model trims pages to the relevant excerpts before they hit Gemini's context window, which is the company's own framing of why the pipe is useful.
The other side has lower friction. Exa Agent shipped as a launch partner on Gemini Enterprise's new Agent Marketplace, alongside entries from Adobe, Salesforce, and Workday. Users in a Gemini Enterprise tenant can add it from the catalog without procurement or code. The agent handles live web research, similar-page lookup, and URL content extraction.
Pricing for grounded calls isn't broken out on the Vertex page beyond a note that customers pay Gemini token costs plus Exa's separate API fees. The marketplace listing rides on Google's A2A agent protocol.
Bottom Line
Grounding with Exa requires an allowlist invite from Google Cloud; the Exa Agent on Gemini Enterprise's marketplace is one click.
Quick Facts
- Two integrations: Grounding with Exa (Vertex AI) and Exa Agent (Gemini Enterprise)
- Vertex grounding status: Restricted Preview, allowlist-gated
- Default rate limit: 200 prompts per minute
- Exa Agent capabilities: web research, similar-page search, URL content extraction
- Other launch-partner agents include Adobe, Salesforce, and Workday
