Working out what is worth doing
Most of the value is in picking the right two or three things to automate, and leaving the rest alone. That part is a conversation, not a piece of software.
Bay Area on-site · remote nationwide
AI systems for your business,
set up properly.
We help businesses and individuals get AI actually working — running locally on hardware you own, or using cloud models with agents built around how you work.
Limited founding spots available $500 personal setup, normally $749
What we actually do
Most of the work is deciding what to build.
Installing the software is the straightforward part. Working out what is worth automating, choosing between local and cloud, and setting it up so it keeps working — that is where the time goes.
Local or cloud, whichever fits
Running models on your own hardware keeps your data in-house and costs nothing per message. Cloud models are more capable and need no hardware. We help you pick, and set up whichever you choose.
Setting it up sensibly
Spending limits so costs stay predictable, access kept to what the work needs, and enough documentation that you can run it without calling us.
If you would rather do it yourself, the notes section covers most of what we would tell you anyway. More on how we work →
Two ways to do this
Local AI, or cloud models with custom agents.
These are genuinely different choices with different trade-offs, and most people are not told there is a choice at all. We set up either, and part of the first conversation is working out which suits you.
Local AI
Models running on hardware you own.
Nothing leaves your machine, there is no per-message cost, and it keeps working if a provider changes their terms. The trade-off is that you need hardware capable of running it, and smaller local models are less capable than the big cloud ones.
Worth considering if
- Work involving client information you would rather not send out
- Predictable costs — hardware once, rather than a monthly bill
- Anyone who wants the thing to keep working regardless of a vendor
Cloud models with custom agents
The strongest available models, with agents built around your work.
You get considerably more capable models with no hardware to buy or maintain, and we build the agents and workflows on top. You pay per use, and your prompts go to the provider — so we are deliberate about what gets sent.
Worth considering if
- Getting started quickly without buying anything
- Work that needs the strongest reasoning available
- Teams that would rather not maintain a machine
Plenty of setups end up as a mix — a local model for anything sensitive, a cloud model for the harder reasoning.
How it works
Three steps, about a week.
One
Strategy session
An hour. We look at how you work, what is worth automating, and whether local or cloud makes more sense for you. You leave with a plan whether or not you build it with us.
Two
Set it up together
We set it up with you — on your hardware or with a cloud model — connect it to your tools, and configure sensible limits. Your machine is backed up before we touch it.
Three
Handoff
A walkthrough so you can run it yourself, plus written notes on how it is set up. Then two weeks to ask anything.
What it costs
Prices, in public.
Everyone else in this category makes you book a call to find out. We would rather you knew now whether we are in your range.
Personal Agent Setup
Most bookedIndividuals and solo professionals
$500 $749
one-time · 2-hour working session
Founding rate — first 10 clients, in exchange for a testimonial and a written case study.
A working AI setup you understand and can run yourself, built around how you actually work.
- Set up on your own hardware, or with a cloud model — your choice
- Connected to the tools you already use
- Spending limits configured, so costs stay predictable
- Access kept to what the work actually needs
- + 3 more
Agent + Workflows
Professionals with repeat workflows worth automating
$1,500–$2,500
one-time
Everything above, plus the agent is taught to do the specific work that eats your week.
- Everything in Personal Agent Setup
- A couple of custom workflows built around your actual work
- Connected to the tools your work already lives in
- Tuned so the running cost stays predictable
- + 1 more
Business Deployment
Teams of 5–50
$5,000–$15,000
project
A setup several people can use, documented well enough that your own team can run it.
- Set up for multiple people rather than one
- Sensible limits on what it can access
- Training for one or two people on your side
- Written guidance on what to keep out of it
- + 3 more
No surprises
What it costs you after we leave.
These bills go to you, not through us. We take no margin on your API spend, and we configure hard caps during setup so it cannot run away.
- Hardware $0 – $799 once
- Reuse a spare machine and this is free. A new Mac Mini runs $499–$799. A VPS is $5–$20 per month instead.
- Model API $10 – $50 / mo
- Personal use at 50–100 messages a day. We configure spend caps during setup so it cannot run away from you.
- Model API — business $50 – $300 / mo
- Team usage across multiple seats and heavier automated workflows.
- Our software $0
- The agent runtimes we deploy are open source. We do not resell them and we take no margin on your API spend.
Not sure which of these you need?
Start with a strategy session. An hour, $125, and you leave knowing what an agent would do for your week, what it would cost to run, and whether it is worth doing yet — including if the honest answer is not now.