Klacks rosters shifts automatically, checks OWTA rest periods and documents planned and worked hours as a basis for Banded Hours applications — open source, installable on your own premises, and optionally with a locally hosted AI model.
No license fees under the AGPL.
Plan a week - Video 1 of 2
Note on the sample data: All names, places (for example Winterthur), addresses and figures in these recordings are fictitious demo data with a Swiss background, written in Latin script. They only illustrate the product – you do not need to know or be able to read them. In your own installation you enter your own data, and the interface is multilingual.
This is how you talk to Klacksy — you can see these sentences in the videos:
One platform, many industries: Wherever staff are scheduled in shifts, in the field, and by qualification — Klacks adapts to you, not the other way around.
You quickly assign a colleague who worked a late shift until 10 p.m. yesterday. Klacks flags it immediately: rest period too short — 9 hours instead of the configured 11. You see the problem before the schedule goes out, not only when someone works it out afterwards.
Watch in the video: rest period too shortShort videos from the running app, sorted by topic. Choose a chapter; only the video of the open chapter ever plays.
Note on the sample data: All names, places (for example Winterthur), addresses and figures in these recordings are fictitious demo data with a Swiss background, written in Latin script. They only illustrate the product – you do not need to know or be able to read them. In your own installation you enter your own data, and the interface is multilingual.
The Republic of Ireland has no public holidays at county or municipal level: all statutory holidays apply uniformly nationwide under the Organisation of Working Time Act 1997. For businesses with ties to Northern Ireland, though, what applies across the border still matters — a completely separate holiday regime is in force there.

Klacks offers countless ways to calculate rules and hour credits for nights, weekends, public holidays and overtime — exactly the way your contracts and company agreements require.
Klacks flags time collisions, rest periods that are too short, too many consecutive working days and work on holidays right as you plan — as an error or warning in the error list.
Rest period too short - Video 1 of 4
From automatic rest-period checks to your own infrastructure — it all fits together.
Klacks monitors the rolling 48-hour average live over the reference period you have configured, as set out in OWTA Section 15, and warns before the limit is exceeded — including the sector-specific extension to 6 months for security and healthcare, if you configure it, correctly factoring in each employee's start date.
On request, Klacks blocks any breach in the automatic planning paths (e.g. wizard, suggestions, substitution); in manual planning, Klacks warns.
The daily rest period of at least 11 consecutive hours (Section 11) and the weekly rest period of at least 24 hours (Section 13), which you set up as rules, are monitored in real time — Klacks makes any deviation visible immediately.
You store the 24-hour advance-notice period for start and finish times required by Section 17 OWTA as the minimum lead time; Klacks uses it to calculate by when the roster must be finished.
There is no automatic check at publication.
Klacks records actual hours worked over 12 months, so an application for Banded Hours under Section 18A OWTA can be answered cleanly within the 4-week deadline.
Klacks runs on your own infrastructure, and you choose the AI model — with a locally hosted model, no AI request containing staff data goes to an external AI provider.
The built-in AI assistant Klacksy understands instructions in everyday language and controls Klacks; the rule-based planning engine calculates the plan — traceable through defined rules, not an opaque AI decision.
Before a plan change goes live, you can run it through as an isolated copy — the 11-hour daily rest period (Section 11) and surcharges are calculated live exactly as in the real plan.
Only once everything checks out do you apply the variant; until then, your live schedule stays untouched. This lets you prepare a suitable scenario in advance for foreseeable situations — such as a staff member calling in sick or a sudden spike in orders — so you're ready to act immediately when it matters, instead of improvising under time pressure. There's no artificial limit on the number of scenarios — create as many as your planning needs.
Klacksy isn't a bolted-on marketing chatbot — it's embedded directly in the Klacks interface. Staff and schedulers ask about shift times via chat or clarify availability. A built-in skills and rules system ensures Klacksy only ever makes transparent suggestions — never a black-box decision. For businesses with strict data protection requirements, the AI model can also be hosted locally, which matters with the DPC in mind too.

Note on the sample data: All names, places (for example Winterthur), addresses and figures in these recordings are fictitious demo data with a Swiss background, written in Latin script. They only illustrate the product – you do not need to know or be able to read them. In your own installation you enter your own data, and the interface is multilingual.
That way staff data is stored on your own servers; the legal assessment towards the Data Protection Commission (DPC) remains with you.
The source code of the backend and the interface is public. Your IT team can check what Klacks does with personnel data — and keep running Klacks independently of us.
Klacks runs on your own servers. Personnel data is held in your infrastructure; the data-protection assessment under the laws of your country remains with you.
Klacksy works with the model of your choice — from external providers to a model you run yourself. You decide on cost, data protection and provider.
Your data lives in your own database (PostgreSQL), the code is open, and you switch the AI provider in the settings.
No license fees does not mean no costs. Servers, operation and maintenance cost money, and so does an external AI model — what you do not pay is a fee per employee.
Your data is stored on your own servers. For address lookup and route calculation, Klacks uses public services based on OpenStreetMap by default; addresses and coordinates are sent there without names. The documentation lists which other external services Klacks calls.
Klacks speaks MCP (Model Context Protocol), the open standard through which AI agents work with software. Any agent that speaks the open standard MCP — such as Claude, Cursor or your own agents — can work directly with Klacks: with the tools Klacksy uses as well, and only with the rights of the connected user.
Operating steps in the interface and sensitive functions such as managing access tokens stay locked, and an agent never receives more than supervisor rights.
How to connect your agentAlso in Klacks:Shift planningRoute optimizationAI assistantScenariosAbsencesRules & surchargesInstalled on your premisesOpen Source
A roster changes at short notice. Klacks checks the 11-hour rest period live; whether the 24-hour advance-notice period under Section 17 OWTA is met is not checked automatically by Klacks — but you no longer have to recalculate the rest period by hand.
Try the Playground to see how Klacks automates OWTA working-time limits, rest periods and Banded Hours — installed on your premises or in the Playground, no license fees and without registration.
Klacks is open source under the AGPL-3.0. Host it yourself, keep your data with you, pay no license fees — you stay in full control.
No demo appointments, no sales calls: try Klacks right in your browser — or install it on your own server or an office PC.
A public Klacks instance with sample data — right in your browser, no installation, no registration.
The installation package contains the installer and the configuration for the database and HTTPS; the installer pulls the Docker images from the registry at startup. One command starts the installation.
Download the installation packageKlacks is open source under the AGPL-3.0. Backend, frontend, and Docker images are public — inspect it, fork it, contribute.
View the repository