My workday starts with three words: "Kick off my day"
Every morning I type "Kick off my day" (Mondays it's "Kick off my week").
That one phrase fires a tool set against my own business data and hands me a visual snapshot of everything in motion: active projects, open tasks, customers, cash position, anything overdue or stuck.
Then the AI shifts into executive assistant mode and asks me four things:
What's the number 1 win for today? What's on your todo list? Any fires? Any updates on the stuck items it just surfaced?
From my answers it structures the rest of the day. It splits the list into what it can do entirely, where it can do the hard part, and what's mine alone. Then we work through it together.
The part that makes this real instead of a chatbot gimmick: the snapshot comes from joins across systems. CRM, ledger, and a persistent memory of past decisions, all read together. That's how it knows a client is 20 days past due AND when I last talked to them AND why we agreed to wait.
The plumbing is Founders OS, an open-source MCP server my cofounder and I built. Runs on your own Supabase over stdio, works with Claude, Cursor, or any MCP client. Your customer list and revenue live in a database you control, not someone else's backend.
GitHub: https://github.com/OurThinkTank/founders-os
Site: https://foundersmcp.com
What does your morning kickoff look like? Curious how other builders start the day.
Our morning is similar but different plumbing. We have agents that pre-process overnight activity across support tickets, inbound leads, and project status before I open my laptop. By the time I sit down there is a triage summary waiting.
Curious about your approach to the data freshness problem. The hardest part for us has been keeping cross-system joins accurate when different data sources update at different cadences. How are you handling that with Founders OS?
Nice setup, the overnight triage layer sounds slick.
We sidestep the freshness problem instead of solving it. Founders OS doesn't federate live queries across separate systems. Everything lives in one Postgres database: CRM, ledger, tasks, memory. The MCP tools read that database at question time, so the joins are always internally consistent. There's no cache to go stale and no cross-source cadence mismatch, because there's only one source.
The tradeoff is that the freshness problem moves upstream: the data is only as current as what you've put in. For a solo founder or small team that's fine, since logging an interaction or a transaction is part of the workflow anyway. If you're syncing in from external systems like a support desk, that ingest cadence becomes your freshness boundary.
Curious how you handle conflicts when two sources disagree about the same entity?
The single Postgres approach makes sense for the consistency win. In our case we work with clients who have fragmented systems so we had to build a different pattern. We use a source-tagged event log where each piece of data keeps its origin and timestamp. When two sources disagree the freshest source wins by default, but anything outside a configurable confidence band gets flagged for review rather than silently merged. That way the system can still operate without perfect alignment. Feels like the decision comes down to whether you can control the upstream or have to accept whatever shape the data arrives in.
Makes sense, thanks for sharing. I like the approach when dealing with multiple data sources that you may not have control over all of them.
The confidence band is the piece I keep thinking about. Flagging for review instead of silently merging keeps the system useful while it is imperfect, which is the realistic bar when you inherit fragmented upstream data. Do you tune that band per client, or does one setting hold up across the board?