Google Deep Research API is live. What would you build on top of it?

This is the combination I keep thinking about:

• Web-off mode: research runs grounded exclusively in private data

• MCP servers: connect custom internal data sources

• Code Execution: run analytics within the research workflow

• API access: call it programmatically from your own application

That combination makes a set of applications buildable that previously required building your own research orchestration from scratch. Clinical literature review. Competitive intelligence with internal data. Legal due diligence. Internal knowledge synthesis that doesn't touch public indexes.

The collaborative planning feature also changes the UX model — instead of submitting a prompt and waiting, you review the research plan and shape it before the system executes. That's a meaningful shift for high-stakes research tasks where the framing matters.

For builders here: are you thinking about this as a primitive to build on, or as an enterprise tool to adopt directly? And has anyone actually tested the Code Execution within a research run — curious how well that works for data-heavy tasks.