What should the standard research layer for AI agents look like?

While building AI agents, I’ve run into a problem that seems bigger than any single agent:

An agent can reason about information much better than it can reliably obtain and verify that information.

Today, solving this usually means stitching together different tools and integrations:

search APIs, website crawlers, browser tools, platform-specific APIs, MCP servers, databases, etc.

That creates another problem: the agent has to know which tool to use, how to use it, and whether the result can

actually be trusted.

For example, imagine an agent receives:

“Research this company and give me its official website, public contact information, current services, recent activity, and sources proving each result.”

Instead of making the agent manually choose several tools, I’ve been exploring the idea of a dedicated research agent that other agents can delegate this task to.

The research agent would:

• understand the requested information

• break it into individual claims

• identify the most relevant source types

• search multiple sources in parallel

• extract only information relevant to the request

• compare sources and detect conflicts

• track when and where each piece of information came from

• return the result together with its evidence

• explicitly report information that could not be verified

The interesting question for me isn't really whether another web scraper is useful.

It's:

Should web research become a standardized capability that agents can delegate to another specialized agent?

And what would you consider necessary before trusting an external research agent with information used by your own agent?

For example, would you care most about source provenance, independent corroboration, freshness, reproducibility, transparent failures, or something else?

I’d especially like to hear from people who are already building or running AI agents and have had to solve web research themselves.