docAnalyzer v2 - Build files, not just answers

Hi Product Hunt. We first launched docAnalyzer here as “chat with your documents.” We’re relaunching today, because that line stopped being true, and the reason is a story I’d rather tell here than on the landing page. Like everyone this past year, we became convinced that agentic is the real shift in AI, and that the part people undersell is the harness around the model: the LLM reasons, but deterministic tools do the exact work, convert the file, build the spreadsheet, assemble the bundle. The model decides what to do; the tools make sure it’s done correctly. We built a new agentic core on that idea. It benchmarked well enough that in March we made the call to retire the old docAnalyzer engine and run on the new one instead. Here’s the part we didn’t plan. The new core delegates work down to a second layer of tools, and buried in that layer was the ability to turn a chat into actual files. We never advertised it. There was no button for it anywhere in the product. People found it anyway, and started coaxing reports and spreadsheets out of a tool that, officially, only chatted. The good kind of jailbreak. That settled it. The product was no longer “chat with your documents,” it was “leave with the work.” So we rebuilt the interface and the positioning around the output instead of the conversation. Here’s what that looks like now. Someone recently dropped 22 sources into a workspace, hundreds of pages of theses, dissertations, and policy reports, and asked one question: write a systematic review. Out of that single request came four files: a cited review in HTML, the same review as a Word document for a committee submission, an Excel table of the studies across four columns, and a single bundle of all three. Every claim traced back to a page in their own PDFs. They never typed “convert,” “build a spreadsheet,” or “bundle these.” They asked a research question. The workspace did the work. What’s in this version: - Chat to deliverable: compose to PDF, Word, Excel, charts, and diagrams, then reuse, convert, or bundle any output in a later turn. - Agentic workflows that run one job across a whole library at once: extract the same fields from every document, audit against a checklist, summarize each. - A new Ask mode that plans and runs work across your whole workspace, alongside Focus and Co-work. - Visibility into what the agent is doing while it works. Underneath all of it, the foundation is the same one we’ve had from the start: source-grounded retrieval with page-level citations, hundreds of documents past the model’s context window, and dozens of models to choose from. It’s built for people judged on the work they ship. The Community plan is free, no credit card, and your first session is ready in about 30 seconds. If you’ve ever pulled an answer out of a chat and then spent the next hour turning it into something you could send, that gap is what we built this for. And if you were one of the people who found the file trick before we admitted it existed: thank you. You basically wrote the roadmap.