Example proposal: review and select web sources

Is your feature request related to a problem? Please describe.

When someone checks the sources behind an answer, they need to compare links and excerpts, then tell the assistant which ones to use. A small app could keep that review in one place instead of making them copy links between tool responses and chat. For developers building it, #437 raises the related setup question: how do you add an App UI around an existing hosted MCP service?

Describe the solution you'd like

Would a small source-review example be useful here? It would show search results, let the user fetch a selected page, and send their chosen sources back as context for the next turn. A small server would own the UI resource and call Parallel's free Search MCP on the server side. That service provides search and fetch tools without an API key, but isn't itself an MCP App. The lesson would be connecting remote tool results to a UI, handling follow-up calls from that UI, and sharing the user's selection through the existing SDK. No SDK or host changes.

Describe alternatives you've considered

Wiki Explorer already demonstrates browsing linked pages. This would focus on comparing search results and choosing evidence for the conversation. A community-maintained example, along the lines of #627, could also be a better home.

Additional context

I'm asking before implementation, per the examples policy. Would this add enough educational value for an in-repo example, or would you prefer the community route? It would make clear that queries and selected URLs go to Parallel, show free-service limits and failures, and use synthetic responses for automated tests so CI wouldn't depend on live search.

Start by reviewing the existing Wiki Explorer example, the examples policy, and the related discussions in #437 and #627. Evaluate whether a source-review example belongs in this repository or a community project, then define the example's scope around Parallel Search MCP, UI follow-up calls, selected-source context, and synthetic tests.