About the project
Klosemore is a conversational sales hub for small sales teams, the three-to-twenty-person kind that are priced out of the enterprise tools and stuck with CRM reporting that cannot keep up. It sits on top of your CRM, starting with Pipedrive, and your inbox, and turns scattered sales data into answers, with no dashboards to build and no reports to wait on.
It is also the product I rebuilt and run solo. The idea started as morr.ai with my co-founders at Ars Ratio, won a pre-accelerator, then stalled, and we stepped back. At the start of 2026 I picked it back up on my own and rebuilt the whole thing from the ground up. The data layer, the detection logic, the AI, the product, all of it is mine.
The approach
CRMs are good at storing deals and bad at telling you what they mean. To get a real answer today, a small team exports to spreadsheets, wires up a BI tool, cross-references email by hand, and rebuilds it every time someone asks something new. The tools that solve this properly, Gong, Clari and the like, start around a hundred dollars per user a month, so a five-person team is priced out. And the signals that matter most, a deal going cold, a champion going quiet, sentiment slipping, slip past because nobody has time to watch.
So I built Klosemore as the intelligence layer between a small team and that mess. Not a CRM, not another dashboard. Three things in one place: it watches the pipeline and flags what matters before you ask, it answers questions in plain English, and it builds the reports people used to assemble by hand.
What I built
All three layers are mine. The first is a signals engine that watches the pipeline around the clock against a baseline of what normal looks like for each workspace, and flags what is worth acting on: a deal stalling in a stage, a champion going dark, a warm reply left unanswered, sentiment shifting, a high-value account going quiet. Each signal explains why it fired and what to do next. The second is the conversational hub, where anyone can ask about their pipeline in plain English, with memory and context that carry across the conversation. The third is a report builder that turns a sentence into a recurring report.
Under the hood it is a full-stack build. Next.js and tRPC, Prisma on Postgres, background jobs on BullMQ and Redis for the syncing and detection, Clerk for auth, and a model-agnostic AI layer that runs across Anthropic, OpenAI, Google and Groq through the Vercel AI SDK. It syncs deals, contacts, activities and email, and writes what it finds back into the workflow. I designed the data model, the detection logic and the product, and shipped it end to end on my own.
Where it stands now
Klosemore is live in private beta, integrated with Pipedrive and email, with more integrations and self-serve signup on the way. It is the clearest proof of what I can build alone now: a real full-stack AI product, taken from a stalled idea to working software by one person, the kind of thing that needed a team and a year not long ago.
I am still building it. More integrations, more detectors, and the path from private beta to an open product are all in progress.
