Open source team check-in

See your team from both sides.

A five-minute check-in each cycle that rates every working relationship from each end, shows how much of your effort reaches the people who judge the result, and asks whether everyone still means the same goal. Open source. Team-level only. No scores on people.

Free for one team on the hosted version. Self-hosted is free for everyone; companies that keep using it buy a one-time licence.

Sample team · rows rate columns

Health checks ask how the team feels. They don’t ask how the team looks from the manager’s seat, or from the client’s. The gap between those views is where most work goes wrong, and nothing you use today measures it.

How it works

Five minutes a cycle.

01

Pick a preset or name your roles

Developers · Management · Stakeholder is one; a platform team and its consumers is another. Any set of roles works.

02

Everyone answers on their phone

Sliders, not scores: how does the relationship look from where you sit? How much of your week reached the stakeholder? Do we still mean the same goal?

03

See it from both sides

Each relationship rated from each end, and the gap between. A quadrant shows effort reaching the stakeholder against the value they say they got.

04

Take three questions into the retro

The AI reads the aggregates and writes questions, never verdicts. You decide what to do with them.

Connections · 1 of 3

How well does your connection with Management work this cycle?

Things get stuck or lostThings flow, nothing gets lost

Can’t judge this one

No numbers while you drag, and “can’t judge” is always an answer. The stakeholder gets a 60-second form, no account.

What you see

Side by side. Never added up.

Developers ↔ Stakeholder
gets stuckflows
34 · 71 gap 37

Two-sided matrix

Every working relationship rated from both ends. The gap is the finding.

Whole team
55%45%
Last cycle
44%56%

Circulation and generation

How much of the team’s energy stays inside, and how much reaches the people who judge the result. Neither is good or bad; the trend is what matters.

S23

Conversion quadrant

Effort toward the stakeholder against the value the stakeholder reports. Watch the dot move from cycle to cycle.

Cooperation network

Who asks whom for help. Team-visible, opt-in, hidden below five people.

Developers62 Management88 Stakeholder70

Goal alignment

Do developers, management and stakeholder still describe the same goal? Where do they want it changed?

StableTransition zone

Timeline and phases

After six sessions: is the team stable, shifting, or in a zone where small things decide the outcome?

Presets

Start from a preset. Change anything.

Dev · Manager · Stakeholder

A delivery team, its managers, and whoever judges the result.

DevelopersManagementStakeholder

Dev · Manager

Product teams whose owner sits in management.

DevelopersManagement

Team · Platform team

A platform team and its consumers.

Platform teamConsuming teams

Founders · Team

Small companies. Nobody outside judges.

FoundersTeam
Experimental

Agents · Humans

AI agents, rated by the people using them.

HumansAI agents

◆ judges the result. Write your own preset →

Privacy

Built so it can’t be used against anyone.

Enforced in the server, not promised in a policy.

  • Results are shown per role group, never per person. If you are the only one in your group, the tool tells you so before you answer.
  • No single “team health” number. No red and green.
  • The cooperation network is opt-in and hidden below five people. A name appears on it only with that person's consent, and only admins see names.
  • No passive data. Nothing is read from your calendar, chat or code.
  • The AI gets aggregates and anonymised text. It writes questions.
  • Self-hosted means your data stays on your server. The hosted version is EU-hosted with a published DPA.
Two ways to run it

Open source. Pay once if it works for you.

Self-hosted

The whole product is Apache-2.0, presets included. Run it for as long as you like. If your company keeps using it, buy a one-time licence.

$900 once, per organisation
docker run -p 8080:8080 -v gyrator:/data ghcr.io/OWNER/gyrator

Unlicensed copies work fully. They print one small line on reports and exports. That’s the entire enforcement:

Hosted

Or let us run it.

Free for one team. $49 per team per month after that; $99 with AI retro questions, a portfolio view across teams and alerts. EU-hosted, SSO, backups, support. Export your data any time.

For agencies and outsourcers

Show your client the same picture you see.

With the Stakeholder preset, the client answers a 60-second form each cycle and sees the pooled report. Renewals stop being a surprise.

Run your first session
Where it comes from

The method came first.

Gyrator began as essays on how teams work as systems. The tool measures what they describe, and what the tool finds goes back into them.

Read the method

FAQ

Is this a performance tool?

No. It cannot produce a per-person result; the product refuses to.

How long does it take?

Under five minutes per person per cycle; the stakeholder form under 90 seconds.

How many people does it need?

It works from two people; the cooperation network switches on at five.

What’s the theory behind circulation and generation?

A systems model of activity by A. N. Malyuta. The method is written up in full and updated as the tool changes; you don’t need it to use the tool.

Can I change the questions?

Yes. Every preset is a YAML file, and there is a visual builder in setup.

Does the AI see names?

No. It receives aggregates and anonymised free text.

Works councils?

There is a one-page explainer for German, Dutch and French consultation; the network feature is off by default.