Skip to main content

​
Overview

Your knowledge base is the source of truth for everything your agent should know about your organization. But users ask questions that reach beyond it: current events, public research, someone’s professional background, a page on a site you don’t own. Without a way to look outward, the agent either falls back to “I don’t have that information” or, worse, guesses. Web Investigator gives your agent a set of trusted tools for reaching into the open web: Google Search, News, Scholar, and Images; page fetching with clean text extraction; GitHub profile lookups; and social profile discovery across LinkedIn, X, Facebook, Instagram, and YouTube. You can turn it on or off for specific agents.

​
What it looks like

A user on your sales-enablement agent asks:
User: “What’s Acme Corp been in the news about this week?” Agent: “Acme Corp announced their Q1 results on Monday, with revenue up 18% year over year. They also unveiled a new partnership with Globex. Here’s a summary from the announcement…” [links to source articles]
The agent ran a news search, read the top results, and composed a grounded answer with live citations. No one updated your knowledge base with this week’s press. The agent just looked. Another example on an L&D agent:
User: “Can you pull a recent study on spaced repetition for onboarding?” Agent: “Here’s a 2025 paper from the Journal of Applied Learning Sciences that looked at spaced repetition effectiveness in corporate onboarding programs…” [cites the paper]
The agent used Scholar to find academic sources, fetched the abstract, and returned a summary with a link. That’s a question your team used to answer by hand.

​
Key features

Web Investigator gives your agent seven capabilities:

Web search

Returns Google web search results with titles, URLs, snippets, and knowledge-graph data.

News search

Pulls Google News results with time filters for day, week, month, or year.

Scholar search

Finds academic papers, authors, and citations through Google Scholar.

Image search

Returns images for a query with URLs, titles, sources, and dimensions.

Fetch webpage

Pulls a specific URL and extracts the readable text content.

GitHub user search

Looks up GitHub profiles and returns bio, public repos, and follower counts.

Social profile search

Finds public profiles on LinkedIn, X, Facebook, Instagram, and YouTube.
The agent chooses which tool to use based on the user’s question. You don’t need to direct it. Ask about a company’s news and it reaches for News search. Ask about a paper and it goes to Scholar. Ask about a person and it reaches for GitHub or the social tools.

​
Use cases

Web Investigator fits naturally into any agent whose users ask questions that go beyond your internal content:
  • Sales-enablement agents that need current news on accounts and prospects
  • Research agents pulling recent academic papers, market data, and third-party reports
  • Support agents referencing public documentation on third-party products your customers also use
  • Internal productivity agents where users expect a “search the web for me” capability inside an agent that already knows company context

​
When to use it (and when not to)

Web Investigator is powerful, which makes it important to be deliberate about where you enable it.
Good fits
  • Sales-enablement agents who need current news on accounts and prospects
  • Research agents that assist with academic or market research
  • Support agents that occasionally need to reference public documentation on third-party products
  • Internal productivity agents where users expect a “search the web for me” capability
Be careful with
  • Agents whose answers must come only from your knowledge base for compliance or accuracy reasons
  • Customer-facing agents where an off-topic web answer could be a support or reputational risk
  • Agents handling sensitive queries where you don’t want the agent reaching beyond your content
A useful rule of thumb: if a wrong or off-topic answer would hurt more than “I don’t know,” leave Web Investigator off for that agent.

​
Things to know

​
The agent decides when to use it

Like any tool, the agent calls Web Investigator when it decides the question calls for it. If you want to nudge the behavior, for example, “prefer the knowledge base, only reach for the web if the knowledge base returns nothing,” say so in your agent’s instructions. The tool description encourages sensible defaults, but your instructions win.

​
Results are live, not cached

Every call hits the live web, so your agent stays current. It also means the same question asked twice may produce slightly different results depending on what the web returns at that moment. This is expected behavior for a live search tool.

​
Citations are returned where possible

For searches, the tool returns source URLs alongside each result. The agent can cite them in its answer, so users can verify or follow up. Coach your agent in its instructions to cite web sources explicitly when it uses them, so users can distinguish knowledge-base answers from web-sourced ones.

​
It doesn’t replace a custom MCP server

If you have a specific external system the agent needs to query (your CRM, an internal search service, a private dataset), Web Investigator is not the right tool. Those belong in a dedicated Connection. Web Investigator is for public web information, not for authenticated integrations with systems you own.

​
Fetched pages are text-only

The fetch webpage tool extracts readable text content. It doesn’t render JavaScript-heavy pages, log into gated sites, or return images or video. If a page is paywalled or requires a login, the agent will get whatever the unauthenticated page returns, which is often a login prompt.

​
Getting started

Turning Web Investigator on is a two-step process: first at app level, then on the specific agent that should use it. Agents you don’t enable it on continue to run exactly as they do today, and your existing setups are unaffected.

​
Prerequisites

You need admin access to Agent Management Studio (AMS) for your app, and at least one agent already configured.
1

Enable the connection at the app level

  1. In AMS, go to Connections.
  2. Open Mindset AI Tools.
  3. Find Web Investigator and turn it on.
This makes Web Investigator available to your app. It does not yet give any agent access to it.
2

Add it to a specific agent

  1. Open the agent in AMS.
  2. Go to MCP Servers.
  3. Type Web Investigator in the search, and turn it on for this agent.
The agent now has access to the tool. No prompt changes or custom configuration are required. The tool is discovered automatically on the agent’s next message.
3

Test it

Try a question that should reach outside your knowledge base, for example “What’s in the news about [a topic your agent wouldn’t normally know]?” Review the agent’s answer and check the cited sources.

​
FAQ

No. The tool is discovered automatically once enabled, and the agent knows how to call it from its description.
Web Investigator is enabled as a set.
Only if you add Web Investigator to them. Existing agents keep running unchanged.
Yes. Any agent configured in AMS can be given access.