THE BLANK PAGE PROBLEM
You already run the agent.
Skip the Tutorial!
Every new agent starts the same way, with an empty file and a blinking cursor. Answer eight questions instead and get four files tuned for the agent you already run.
THE COST
What the blank page costs you.
Hours you do not get back
A usable agent file is an afternoon of writing, testing, and rewriting. Most people do it twice, because the first pass was a guess.
Agents that do not match
Two agents set up a month apart answer the same question two different ways. Neither one is wrong. Nothing lines up either.
Drift
The rules you set in January are gone by March, buried under corrections nobody wrote down.
HOW IT WORKS
Three steps to a working agent.
Answer
Eight questions at the front door. Your business, your customers, and one ordinary moment your agent should handle.
We assemble
Your answers land on a base built for your agent class, tuned by a dialect profile from our own testing lab.
You deploy
Four files come out, ready to use. Drop them into the agent you already run.
THE DELIVERABLE
Four files. That is the whole product.
audience.md
Audience
Who your agent talks to, what they care about, and the words to never use.
design.md
Design
Your brand as law: colors, type, logo, imagery, as locked tokens.
agent-instructions.md
Agent Instructions
Role, core behaviors, hard limits, and when to hand off to a human.
voice-tone.md
Voice & Tone
How your agent sounds, from greeting to complaint to close.
PROOF
Works with the agent you already run.
Lab graded
Graded in our own lab, mission by mission.
Supported
The format fits. Not yet graded.
PRICING
Simple pricing. Serious files.
SOLO
$24/mo
or $240/yr
Single user, Learning Files, 10 agent builds a month, project build files unlimited. No product updates: Solo improves through its own learning files.
STUDIO
$59/mo
or $590/yr
3 seats, one shared lawbook, Locked Standard, all updates included, builds unlimited.
ENTERPRISE
$199/mo
or $1,990/yr
5 seats, up to three lawbooks, updates first as the lab grades them, direct line to the lab.
Custom Enterprise
More than 5 seats, or BYOK, local models, and staged output shapes. Negotiated with the lab.
STRAIGHT ANSWERS
Three things people ask first.
Do I need to code?
No. You answer eight questions in plain English and four Markdown files come back. If you can fill out a form, you can run this.
Do I need an agent first?
Yes. AgentBrain tunes an agent you already run, it does not host one for you. If you have Claude Code or Hermes open right now, you are ready.
What if my harness is not listed?
The files are plain Markdown, so they work anywhere that reads instructions. Lab graded means we tested that harness mission by mission. Supported means the format fits and we have not graded it yet.
The early list
AgentBrain is in the lab, being graded the hard way. Leave an email and you are first in when the prototype cohort opens.
One email when your seat opens. No spam, no drip campaign.
If it turns out not to be for you, one click unsubscribes and we will leave you alone.