Asked, not mined.
The judgment that never generated a ticket, email, or transcript is invisible to every tool that mines exhaust. Brevity reaches it the way a chief of staff would — by asking.
The operating loop for agentic work
The half of your company that was never written down. Working.
Brevity learns how work actually happens — by asking, not by watching — and turns it into agents that run under your checks. It stays yours. It gets better every cycle.
The missing half
Your company knows more than its systems do.
Every tool that promises to understand your organization reads the same sources — tickets, wikis, transcripts, dashboards. The exhaust. But the judgment that calls a rollback at 2 a.m., the unwritten order of a release, the reason nobody deploys on the Tuesday before quarter-close — none of it ever generated a ticket. It lives in people's heads. And it's the half that runs the company.
Documented-vs-actual process gaps of roughly 50% have been reported in enterprise process audits.
You can't mine what was never written down. You can ask.
The loop
Captured by asking. Better every cycle.
-
01
Captured by asking
Brevity interviews your people the way a chief of staff would: in conversation, while they work. Twenty minutes reaches what ten years of tickets never recorded.
-
02
Compiled into skills
What it hears becomes a runnable skill. Versioned, reviewable, in a file you own.
-
03
Run under your checks
Agents execute it under the gates you already trust: CI, tests, code review. Nothing merges without evidence.
-
04
Verified against outcomes
Every run is checked against what actually happened. The audit trail writes itself.
-
05
Refined every cycle
Each execution teaches the skill. The procedure your best engineer described in April is sharper by June — even after she changes teams.
Why it holds
Three things nobody else can say.
Yours, not locked.
What it captures comes back as a portable skill you own. A file in your repo — not a graph you're trapped in.
The lane is theirs.
Everyone who teaches the system keeps a lane that is provably their own. People tell the truth to a system that's theirs — that honesty is what makes the capture good.
One graph, three verbs
One graph of how your company works. Three ways it works for you.
Understands
Reverie
Assembles your organization's operating picture ambiently, from conversations — where the work is actually decided.
Manages
Helper
One assistant per person. Right-sized alerts, decision points that genuinely need you — and nothing that doesn't.
Executes
Hox
Governed agent fleets that ship real work under your gates. Every action scoped, evidenced, reversible.
Under the hood, all three share one substrate: a graph where one relation binds many things — the person, the skill, the check, the agent — into a single connected fact. Yes, a hypergraph. That's the only time we'll say it.
For engineering leaders
Your agents inherit how your best engineers actually work.
Verify is free.
CI, tests, and review gates are already there and already trusted. Brevity's agents merge only through them — no new process to invent.
Audit trail included.
Who did what, under which check, with whose approval. Every action carries its evidence.
Scopes, not keys.
Agents act inside granted scopes with your existing secrets hygiene. Blast radius is a setting, not a hope.
Better every sprint.
Skills refine against real outcomes in your repos — not benchmarks.
Built first for teams of 50–500 engineers already running coding-agent fleets.
Ops or regulated? The same loop applies wherever work has checks. Talk to us.
Portable by design
Skills are files you own. Leave any time, take everything.
Private lanes
Every contributor's lane is cryptographically theirs.
Full audit log
Every capture, every run, every approval. Exportable.
No model lock-in
Runs across your agents, harnesses, and models.
It learns from you by asking. The lane is yours.
Early access is open for engineering teams running agent fleets.
Preview build — submissions aren't stored yet.