base - The system of intelligence for lean founding teams
by•
Base gives lean founding teams the operating layer of a much larger company. It is a command interface and system of intelligence, five layers:
Setup, Orchestration, Company Brain, Memory, Intelligence.
Instruct it in natural language to pull financial reports or draft investor updates, and it executes across your connected stack, getting sharper the longer you use it.
Vibe coding got you to a product. Base gets you to a business.
Replies
Hi Product Hunt,
I'm Caroline, founder of Base.
I've spent the last decade scaling four venture-backed businesses, three reaching unicorn valuation, and investing at a private equity fund. Sat on both sides of the table as an operator and an investor, the same thing kept showing up. Founders can ship a product in a weekend now, Lovable, Cursor and Bolt took care of that. Building the business behind it is a different problem entirely, and there's no equivalent shortcut for contracts, CRM, hiring and finance all landing at once with no team to handle them.
So I built the thing I wish someone had handed me. Base gives lean founding teams the operating layer of a much larger company. Instruct it in natural language to pull a report, draft an update, chase an invoice, and it executes across your connected stack.
Underneath sits a company brain built on a knowledge graph. Every contract, invoice, deal and conversation gets resolved into entities and relationships rather than sitting as isolated records in separate tools. That's what makes the difference between a tool that executes one command and a system that knows your business is behind on a payment, which deal's gone quiet, and what actually needs you today, without being asked. It's the part of Base that compounds. The longer you use it, the sharper it gets.
Come connect a tool or two, run it for a day, and tell me where it breaks. I'm reading every comment today.
Caroline
@caroline_hazlehurst really interesting launch🙌 vision behind Base is ambitious in the best way.
@priya_kushwaha1 thanks so much for the support !
qq what underlying graph architecture keeps query latency sub-second when re-indexing complex entity relationships across multiple SaaS tools? Congrats for shipping @caroline_hazlehurst 👏
@vikramp7470 Great question. Today it's deterministic entity resolution on canonical tables rather than a graph database, a deliberate choice while volume is still building, since a graph layer earns its complexity cost at a different scale than where we are now. Per-adapter caching with webhook-triggered invalidation means most updates only touch the one tool that changed, rather than a full re-index. The architecture we're building towards is agent long-term memory with consolidation, then cross-tenant learning over time. Happy to go deeper, always good to talk to people who ask this kind of question.
@caroline_hazlehurst Solid explanation, appreciate the pragmatic approach decisions like these make all the difference.
@vikramp7470 appreciate the support !
@caroline_hazlehurst Congrats on the launch! Curious how far back it can look once you connect a tool, does it work with historical data or only from the day you connect it?
@abhiranjan_mehta Yes, it works with what's already there once you connect a tool, not just from the day you connect it. That's actually the more valuable case, most founders come to this with some messy data already, not a blank slate
This is exactly the gap I hit last year. No CRM, no idea who to chase for payment, just vibes and a spreadsheet. Wish this existed then.
@himanshi_kum563 I should have built this earlier haha!
@himanshi_kum563 i should have built this earlier!
@caroline_hazlehurst This is so needed. I’ve watched three friends ship a product then completely stall on the business side. Bookmarking to send them this.
@maurya_abhiranjan Thanks so much!!! That's exactly who we built this for. Would love an intro if they're up for trying it, always better to hear where it breaks from someone in exactly that spot
@caroline_hazlehurst How does this handle a founder who’s already got messy data across five tools before connecting Base? Or is day one meant to be a clean slate?
@suyash_kr Good qu!! day one isn't a clean slate requirement. Base pulls in what's already there when you connect a tool, so existing history feeds the picture rather than starting fresh. It gets more useful the messier that history actually is, since that's exactly the kind of thing it's meant to untangle for founders !
This looks awesome!! Exactly what I've been wanting to open up every morning.
@gpat_uk Thank you!! Appreciate the support and would love to hear how it goes once you've had a proper run with it !
"System of intelligence for lean founding teams" is a broad promise — curious what that looks like concretely day-to-day. Is it more about surfacing insights from existing tools, or does it actively suggest next actions?
@ark_y_k Bit of both, but it leans toward the second. It'll flag what's overdue or gone quiet without you asking, and increasingly point at what to do about it rather than just surfacing the data point.
@caroline_hazlehurst Proactively flagging what's overdue or gone quiet — and increasingly suggesting what to do about it — sounds like it's aiming to be more of a co-pilot than a dashboard. That shift from "here's the data" to "here's what to do" is probably where most of the real value ends up living.
@ark_y_k that's exactly right, so pleased to hear you get it !
@caroline_hazlehurst Glad I got it right! Sounds like you've got a clear sense of where this needs to go, which is half the battle at this stage.
Dial
the knowledge graph piece is the ambitious part - entity resolution across contracts, deals and conversations is genuinely hard, and it's exactly the kind of thing that looks perfect in a demo and then merges two different companies with similar names into one entity. when it's wrong about something like that and 'chases an invoice' based on the bad merge, does a founder find out before it sends anything, or after?
@omri_ben_shoham1 Base is human-directed, not autonomous, so it drafts the invoice chase and waits for you to approve before anything sends, a bad merge shows up as a wrong name in front of you, not a wrong email in someone’s inbox.
Dial
@Caroline that's a good distinction actually - a wrong name sitting in a draft you review is a completely different risk profile than a wrong email already sent. thanks for clarifying, that human-in-the-loop framing makes a lot more sense for something touching invoices.