Fixed scope. Fixed price. Agreed before we build.
There's no menu of packages to choose from. Every engagement starts with the same week-one function map, and that map can end things before you have spent the rest.
Pricing
$0is the floor, not a package.
There's no menu of tiers to pick from. We map the work in week one, then quote a real number for exactly what you need — from a single agent to a full embedded team.
Book a callTypically live in under a month.
What moves it from there
Systems touched
How many tools, databases and platforms the agent has to work inside from day one.
Compliance line
How strict the guardrails need to be before the agent is allowed to touch anything unsupervised.
Data condition
How messy or well-kept the underlying data already is before the agent has to read it.
Model and infrastructure costs are quoted separately, and are almost never the deciding factor.
Regardless of what we end up scoping.
An honest go / no-go
At the end of week one, in writing. We have ended engagements here and refunded the difference.
Guardrails you define
What the agent may do unsupervised is your decision, written down before the build starts.
Your code, your repo
Prompts, evaluation sets and documentation included. No platform fee, nothing held hostage.
One metric, reported
Agreed before we start, measured after launch, reviewed monthly. No vanity dashboards.
The ones we get asked on every call.
Below that, the discovery alone consumes the budget and you end up with a demo rather than something in production. Three thousand is the point where a scoped agent can be built, evaluated and put in front of real traffic properly.
Three things, in order: how many systems the agent has to touch, how strict the compliance boundary is, and how messy the data behind it is. Model costs are rarely the deciding factor. We quote after the function map, not before.
Build fees are separate from running costs. We size the expected monthly inference and hosting spend during scoping and show you the arithmetic. Most single agents run in the low hundreds of dollars a month.
You do. Code, prompts, evaluation sets and documentation are yours, in your repository, from the first commit. There is no platform fee and nothing is held hostage if you leave.
Shadow mode exists so that you find out before your customers do. If the agent cannot clear the metric we agreed on during evaluation, we do not put it live. That is also why we run a written go / no-go at the end of week one.
Often. We can build the agent and hand it over complete, or work alongside your engineers on a rolling basis so they own it from day one — whichever the work actually needs.
We quote a real number after the week-one function map — the $3,000 floor is the only figure fixed in advance.
Get a number for your actual situation.
Pricing depends on how many systems the agent touches, how strict the compliance boundary is, and how messy the data is. Thirty minutes gets you a real range.
Engagements start at $3,000. Typically live in under a month.