
Substrate
Help Claude understand your customers.
AI & Machine Learning·Productivity
What is Substrate?
Help Claude understand your customers.
Stop re-explaining every customer to Claude. Substrate keeps a page on each one, current from your Gmail, Notion and files, and Claude Code, Cowork and Codex recall it on every prompt.
Read every page. Edit any line. Delete a line and Claude stops seeing it.
I run my own company with Claude every day. Emails, follow-ups, prep before I talk to a customer.
And every time, I was pasting the same context in again. Who this customer is. Who I talk to there. What they asked for last month. What I still owe them.
Claude is very good at the work. It just doesn't know your customers unless you tell it again.
So I built Substrate.
Drop in a customer file and watch their page appear.
-
Connect Gmail or Notion, or upload a file. For Gmail, you pick how far back it goes. Newsletters and automated mail are skipped.
-
Substrate writes one page per customer - the organization, the people there and the open tasks.
-
In Claude Code, Cowork or Codex, the relevant facts come with each prompt. You don't paste anything.
-
You read the page, fix a line or delete it. Every line cites the source it came from.
A small example. You have a quarterly review with a customer next week, so you ask Claude to prepare the deck.
It already knows who will be in the room and what you still owe them. First draft in one prompt - you don't paste anything.
If a line is wrong, you fix it on the page. If you delete it, Claude stops seeing it.
Some honest limits.
Recall brings the facts that are relevant to your prompt, not everything on the page. That is on purpose - it keeps the context small.
Gmail goes only as far back as you choose. If you pick the last 90 days, older threads are not there.
And it's an open beta, so some things will be rough.
Sign-in is OAuth. No API keys to copy around.
It's for founders and operators who already do a lot of their work through Claude. If you explain the same customers to it every week, this should save you that.
First 2 weeks free with code LAUNCH26, then $20/mo.
About me. I'm Pavel. I spent 10 years at Hilti, first as an engineer, then as a commercial leader. Then I built Sitescope, a construction AI start-up. I'm an MIT Sloan Fellow, based in Boston.
Now I'm a solo founder. I build with agents every day, and I run my own company on Substrate.
Would love your feedback. What's missing on a customer page for you? What's confusing in the first five minutes?
Ready to launch?
Submit your product and get discovered by builders and creators worldwide.
Launch NowFounderPlaybooks.
What other founders did to grow.
2722 dispatches from hundreds of founders, pulled from the week's best podcasts.
JJ Allaire was tracking his calories in an Excel sheet and thought 'well I could be doing this better on this new fancy handheld device called the iPhone' and he built the very first version of Lose It, put it up on the App Store. The email inbox started getting flooded with people saying 'this app is amazing it's helping me lose weight' — kind of realized like oh I have a product, I have a company here.
Excel-tracking founder shipped v1 — inbox flooded with testimonials proved demand
Lose It! founder JJ Allaire scratched his own itch by replacing an Excel calorie tracker with an iPhone app at the dawn of the App Store. The contact email he was forced to provide for submission immediately filled with unsolicited 'I'm losing weight' testimonials — that's how he realized he had a product worth building a company around.
We're not looking for apps that aren't making any money that's not great we're looking for pre-validated ideas so increase that asking price to at least 300,000 or you can filter by annual recurring revenue.
Filter business broker listings for pre-validated ideas worth copying
Adrian's framework turns Acquire.com into a validation machine: if a SaaS is listed for $300K+, real money has already been proven in that market. Rather than testing a hypothesis, you're cloning a demonstrated outcome — cutting years off the validation phase.
There's a play for whatever you're stuck on.
Read all 2722 playbooks