Bring one customer's context.
Leave with a governed brain.
Your customer's context, governed, in half an hour
No deck. We use your real context, and everything below is software that exists today — you'll be clicking it, not watching a video of it.
- 010–5 min
What's your brain unit?
One per customer, per workspace, per project? Most people answer 'one' and then work it out mid-sentence. We also cover what you built yourself before you could ship — ingestion, history, deletion — and what happens today when a customer says your agent got something wrong.
- 025–20 min
We stand one up
One customer's context goes in as a brain. We publish it over MCP with scoped access, connect an agent, and it proposes a change back — carrying which agent, which version, which model, which run. Nothing it proposes is live yet.
- 0320–30 min
Reject one, keep the rest, take it with you
You review the proposal item by item, reject one and adopt the others, then clone the brain the way you'd clone it for customer number two. We export the whole thing as a portable bundle before we hang up, so you can see for yourself that nothing is captive.
Bring the question you'd use to kill this in a weekend — "why wouldn't I just build it on Postgres?" is the right one, and I'd rather answer it in minute three than minute twenty-nine.
Is this actually worth your time?
I'd rather you skip the call than sit through one that wasn't for you. Here's the honest read.
Book it if
- Your agents serve many customers, and each customer needs their own context
- You built per-customer ingestion, history, and debugging yourself before you could ship the actual product
- A customer told you your agent said something wrong, and you couldn't prove what it read
- You're onboarding customer number twenty and the per-customer setup is still manual
- Someone will ask how one customer's data stays separate from another's before this touches production
Skip it if
- One shared knowledge base covers all your customers — you don't have a fleet problem yet, and I'd rather you know that than pay to find out
- You're shopping for a chatbot over your docs
- You want me to tell you which vector database to pick — genuinely not the interesting part
- You want one workflow rebuilt as an agent your team can run — that's the other session, and it's the better use of your 30 minutes — See the workflow teardown
- You need paperwork before a working session (email me and we'll start there instead)
Not sure which column you're in? Tell me what your brain unit is today in two sentences: paul@nicia.ai
I live in Bend, Oregon. Seattle is the city I keep coming back to — close enough to drive, far enough that I still notice how good the coffee is. Fair warning: I'm from a town where strangers make eye contact and start conversations. I'm told that isn't how it's done up here. I'm going to do it anyway.
You get me for the whole 30 minutes — no SDR, no handoff, no "let me loop in a solutions engineer." I'm the one who writes the code, so if the thing you need doesn't exist yet, you'll get a straight answer about that too.
- Bend, ORwhere the laptop lives
- Seattle, WAhome away from home
~330 miles · about six hours · one podcast about databases
What we're building
Models are good enough. The missing piece is the operating layer between "works for me" and "works for the team": shared state, policy, evidence, and improvement that a human signs off on.
- Nicia Platform
- Runs your agents for the team — budgets, policy, approvals, audit, and evaluation-driven improvement you approve before it ships.
- TypeGraph
- The typed knowledge graph underneath, so agents read structure instead of guessing from a pile of text.
- Corpus
- Versioned context your non-engineers can actually edit, served to agents over MCP.
Coffee's on me
Anywhere except the original Starbucks. That's a forty-minute line for a cup you can get, better and sitting down, two blocks up the hill.
Weather is not a problem
If it's raining, nothing about this changes. If the mountain is out, the whole city goes outside and we get the place to ourselves.
One thing I won't build
Your agent is not writing LinkedIn posts. It could. The technology is right there. Please don't make me part of that.
The customer you onboard next
shouldn't need new infrastructure.
Grab whatever slot suits you. I'm around Seattle this week and on a calendar the rest of the time.
No charge, no deck, no follow-up sequence.
Calendar full or you'd rather write first? paul@nicia.ai