From a question to a cited report: Deep Research in JoeBro

I’m Joe, the creator of JoeBro. I wanted research to leave behind something more useful than a long chat transcript. JoeBro’s Deep Research takes a question, reads multiple sources, and produces a structured report with citations and relevant images.

You can start a run from the Deep Research search box, or ask for research in a chat and let the agent use the research tool. A run continues if you switch tabs; the interface shows when it is working and when a report is ready. Past reports can be reopened, copied to the clipboard, downloaded as Markdown, or deleted. That gives the result a life outside the conversation: you can keep it with project notes, revise it, or share the Markdown through your usual process.

Citations make it possible to follow the report back to source material, but they are not a substitute for checking the sources. A useful review asks whether the cited page actually supports the sentence, whether the source is current enough for the question, and what the report may have missed. The feature is a research aid, not a guarantee that every source is correct or that a report is complete.

Privacy depends on the endpoints involved. JoeBro’s workspace data is stored locally, and you choose the model endpoint. If you choose a hosted model API, prompt content sent for that task goes to the provider for processing. Search requests also involve the configured search service. Cloak Mode runs before requests to model and search endpoints, including local ones, and replaces values it recognizes as sensitive. It is pattern-based, so it can miss unusual formats; it should not be treated as a promise that no sensitive information can leave your Mac. Local inference changes the model boundary, while web research still requires contact with web sources.

I’m Joe, creator of JoeBro for macOS and JoeBro Mobile. The phone app is a companion to the Mac workspace, not a separate research service.