Free engine. Managed platform when you want it.
The workflow engine is open source and always free. The managed platform, where you stop provisioning infrastructure entirely, opens in January 2027. Prices are announced at launch, and waitlist members get them first along with a launch discount.
Open source
€0
Available now
Anyone who wants to run workflows on infrastructure they already have.
- horus-runtime, the full execution engine
- Self-hosted, CLI and Python API
- HPC schedulers, cloud and local execution
- Community support, AGPL v3
Starter
Coming soon
Launching January 2027
Individual researchers and small groups with no platform engineer.
- Managed compute, nothing to provision
- Automatic auto-scaling
- Visual workflow builder and live monitoring
- Included monthly compute, top up on demand
- Standard support
Pro
Most teamsComing soon
Launching January 2027
R&D teams of roughly 5 to 15, where this replaces dedicated DevOps work.
- Everything in Starter
- Multi-user collaboration and shared workspaces
- Parallel workflow branching and fan-out
- Full plugin catalog
- Priority support
Enterprise
Let's talk
Pilots open now
Pharma, large labs and HPC centres with their own hardware and rules.
- Everything in Pro
- On-premise and hybrid deployment
- Contractual uptime guarantees
- SSO, audit trail and usage reporting
- Custom domains and dedicated support
One unit, not a cloud bill
Compute is billed in tokens. A token rolls infrastructure, software licensing, storage and data transfer into a single number, so the cost of a run is something you can predict before you start it. One token buys an hour on a standard 16-vCPU node; bigger CPU nodes and GPUs cost proportionally more. Subscriptions include a monthly allowance, extra tokens are bought on demand, and the rate improves at higher tiers.
Raw per-hour cloud pricing looks cheaper on a spreadsheet, and it is not what teams actually pay. Where the money actually goes.
Questions
- Why are there no prices on this page?
- Because the paid tiers are not open yet. Publishing a number before the first customer is billed would be a commitment we are not ready to make, and we would rather say so than show a figure that changes. Waitlist members hear the prices first, and get a launch discount.
- How is compute billed?
- Through compute tokens, so one unit covers infrastructure, software licensing, storage and data transfer together. One token buys an hour on a standard 16-vCPU node; larger CPU and GPU nodes cost proportionally more per hour. Each subscription includes a monthly allowance and you can top up on demand. The point is that there is no separate cloud bill arriving behind the subscription.
- Do I need to pay to use Horus at all?
- No. horus-runtime is open source under AGPL v3 and always will be. If you already have a cluster or a cloud account and someone to look after it, that is a complete way to use Horus. The paid tiers exist for teams who would rather not run the infrastructure themselves.
- Can I run this on our own cluster?
- Yes, on the Enterprise tier, including hybrid setups that burst from your own hardware into managed cloud capacity. Nothing is locked to a single cloud provider.
- What happens to my workflows if we stop paying?
- Workflow definitions export as self-contained archives, and the engine that runs them is open source. There is no format you cannot take with you.
Hear the prices first
Waitlist members get launch pricing and a discount that holds.
Want an account now instead? Request a beta invitation.