Know if AI is working
across your team.
Adoption, spend and effectiveness, read straight from a private git repo you own. Your data never touches our cloud.
Adoption, spend and effectiveness, rolled up.
Every number is a team-level aggregate. There is no per-person leaderboard and no way to drill into one dev — the what-works detail stays on each seat, below.
- effect-kit
- ship-checklist
- ax-extract-workflow
- acme-webjoined
- acme-billingjoined
- acme-infranot joined
This is the live ax studio team rollup embedded on sample data — the same dashboard your cohort gets, not a screenshot. The per-seat surface already renders these numbers for one dev today.
It still tells you what works — on every seat.
The rollup rides on the same engine every dev already runs locally. The full-detail answers never leave the laptop; the team layer only reports whether the wins spread.
- effect-kit
- ship-checklist
- “plan, then edit”
Each dev’s local studio renders these live today — numbers here illustrative. The rollup above is the same answers, aggregated across seats.
Author once, runs on every laptop.
The mechanism that turns one engineer’s trick into team practice already ships: commit a skill or a typed Effect hook and everyone’s agent runs it; the improve loop mines repeated mistakes into reviewed fixes; public profiles and /leaders already prove an aggregates-only rollup works. The governed team registry on top is what we build with the founding cohort.
ships today
- Skills & hooks SDK (commit once, every agent runs it)
- The improve loop (repeated mistakes into a reviewed fix)
- Public profiles &
/leaders(aggregates only, no code sent)
building with the cohort
- Sync down (blessed skills read-only on every laptop)
- Suggest up (agents send a PII-redacted repro upstream)
- Review queue (ranked by real usage; your engineers approve)
Fix flowed back: clause-numbering edge case
- seats hit
- 9
- failures, 14d
- 23
- repro
- synthetic, PII-free
- matter
- never sent
Mock of the review queue we are building. Numbers illustrative.
Consumers’ agents propose; your engineers approve. Regulated mode adds an adversarial recover-pass, a local consent gate and a revocable provenance stamp before anything leaves. Built for privilege, PII and compliance.
The dashboard our servers never see.
ax’s servers never receive or store your telemetry, transcripts, source, prompts, or derived rows. The frontend reads your data with the viewer’s own GitHub token. The only service we run is a stateless auth broker.
Aggregates land in a private repo in your own GitHub org, one redacted file per dev. GitHub repo membership is team access.
The dashboard aggregates client-side using the viewer’s own token. If we get breached, there is nothing of yours to leak.
Nothing pushes until a dev runs ax team join inside a specific repo. Repo identity is pinned, so a fork or rename can’t leak the wrong one.
The entire file that ships.
Names, not contents. No transcripts, no code, no prompts, no paths. Daily-collapsed counts, sums and ratios.
{
"login": "necmttn",
"window": "2026-06-01/2026-06-30",
"sessions": 214,
"active_days": 22,
"tokens": { "in": 4120000, "out": 385000 },
"routable_usd": 605,
"spend_usd": 2140,
"top_skills": ["effect-kit", "ship-checklist", "ax-extract-workflow"],
"churn_episodes": 7
}This is the whole file that leaves. Names, not contents.
Aggregate, never surveillance.
Team cells with fewer than 5 contributors are hidden. No individual drilldown. No per-person leaderboard.
Setup is minutes. It is just git.
No agents on your infra. No transcripts leaving machines.
1 · A private repo
Create a private ax-team repo in your GitHub org and add your devs. Repo membership is team membership.
2 · Each dev opts in
Inside a work repo: ax team join <org>. A consent screen shows exactly what is shared. Personal repos are never joined.
3 · Open the dashboard
Log in with GitHub. It reads the repo with your own token and renders. Aggregation happens in your browser.
The questions your security team asks.
Where does the data live?
In a private git repo in your own GitHub org. One redacted file per dev. We keep no copy.
Who can read it?
Members of that repo, using their own GitHub token in the browser. Repo membership is the only access control.
What actually leaves a machine?
Only the redacted aggregate file shown above. No transcripts, code, prompts or paths.
What if a dev leaves or a token leaks?
Remove them from the repo and access ends. A leaked token exposes only aggregate counts, never source or transcripts.
GDPR and deletion?
Delete the dev’s file from the repo. That removes their data everywhere, because the repo is the only store.
Self-host or exit?
ax is AGPL-3.0. The data is already yours in git, so there is nothing to export and no lock-in.
Five teams, then we close it.
The rollup is real work built around real repos. We can hand-onboard and hold a weekly founder call for five teams this quarter, no more.
1 claimed · 4 open
- $20/seat locked. Founding price holds as the product grows.
- Direct founder line. Onboarding by hand, weekly call.
- Shape the roadmap. We build the rollup around your repos.
$20 per developer, per month.
Per-seat, self-serve, cancel anytime. A seat is a dev who pushes. It rides as a small add-on to the agent subscriptions you already pay for, and pays for itself the moment it redirects one routine sub-task off the expensive default.
You pay for the dashboard and the enablement. Your data stays in your git repo; we store none of it.
Visibility without the data-grab.
See it on your own local ax data today, then decide if the team layer is worth piloting.