The code is free. The framework and its ready-made building blocks are BSL-1.1 on GitHub, and self-hosting costs €0, permanently. This page gives guide values for kumiko.rocks Hosted, the managed platform that runs your apps for you.
Status: working draft. All amounts on this page are guide values, not final prices. Final prices will not come before Q1 2027.
What’s already settled
- The framework and building blocks are public: BSL-1.1 on GitHub (
CosmicDriftGameStudio/kumiko-framework), and the NPM packages are free to use. After 2 years, every release automatically becomes Apache-2.0 - The platform code (Designer, AI builder, setting up new customer environments, billing) stays in a private repository. That protects us against cloud clones
- kumiko.rocks Hosted comes in plans
- Your own AI key (BYOK, “bring your own key”) is the default, not a premium extra. Your running costs for AI, email and storage belong in your accounts (see below)
- There is no per-user pricing, unlike Retool. At 100 employees you don’t pay 100 times as much
Plans at a glance (guide values)
| Plan | Price range (guide value) | Apps | Hosting | AI | Support | Audience |
|---|---|---|---|---|---|---|
| Self-Host (Open Source) | €0 | n | your choice | own key (BYOK) | Community | self-hosters, mid-market with its own IT |
| Free (Hosted) | €0 | 1 | kumiko.rocks/sub | Anthropic demo with a cap | Community | indie hackers, MVPs |
| Pro (Hosted) | from €49/month per app | 5 | kumiko.rocks/sub or own domain | own key (BYOK) | solo developers, small teams | |
| Business (Hosted) | on request | 20 | like Pro, with higher caps | own key (BYOK) | Email, priority | growing teams, multiple apps |
| Enterprise | mid 4–5 figures p.a. | per contract | dedicated or on your own premises | local model or own key | SLA and premium support | European mid-market, regulated industries |
The price range for Pro and Business applies per app, not per user. Apps with several tenants (separate customer areas) run on one instance.
Plan changes without data loss
When you move to a bigger plan, your data stays where it is. We only change the resources your app runs on, and the database is left untouched. The whole change takes about as long as a restart of a few seconds. You don’t have to export and re-import anything, and there is no maintenance window that turns into a maintenance weekend.
Going back down works the same way. If you booked too much, you scale down without losing anything.
Many platforms for internal tools turn a plan change into a sales conversation, because the migration is painful for them. For us it is unremarkable on purpose, because that is the safer path.
Running costs belong in your accounts
Whatever costs money depending on usage runs through your accounts, not ours. We don’t use a shared platform key as a business model. This has three consequences:
- You keep your reputation and your control. When you send email to your customers, it goes through your Brevo or Postmark account, not ours. Spam problems and reputation damage hit you, not 50 other tenants on the platform
- Your real costs become visible: the Anthropic bill at Anthropic, the S3 bill at Hetzner, the SMTP bill at Brevo. The plan covers running the platform, not your external services
- It helps with compliance: “your data, your AI and your mail never leave your accounts” uses the same architectural lever as running on your own premises does for European buyers
| Service | Default from Pro onwards | Free, for trying out |
|---|---|---|
| AI tokens | own key required (Anthropic, OpenAI or Ollama) | demo with 100k tokens through the platform key |
| Email sending | own account required (Brevo, Postmark or SES SMTP) | 1,000 platform mails per month (login, password reset, notifications), no mass mailing |
| File storage | own S3 storage (Hetzner Object Storage, MinIO or AWS S3) | 1 to 10 GB on the platform, then your own storage |
Caps per plan
We cap platform resources, not your business logic. What your app does internally with tenants, workflows or data structure is yours. We only measure what flows into our Postgres database, our mail server and our AI compute (GPU pool).
| Cap | Free | Pro | Business | Enterprise |
|---|---|---|---|---|
| Apps | 1 | 5 | 20 | per contract |
| Database storage (summed across all apps and internal tenants) | 100 MB | 10 GB | 100 GB | per contract |
| File storage (platform) | 1 GB | 10 GB or own S3 storage | 100 GB or own S3 storage | own S3 storage |
| Platform mails per month (login, password reset) | 1,000 | 10,000 or own SMTP | own SMTP required | own SMTP |
| AI tokens (rolling 7 days) | 0 (no AI Designer on the Free plan) | 50,000 | 150,000 | own key or local model |
| Outbound traffic (egress) | 10 GB per month | 100 GB per month | 500 GB per month | per contract |
How the caps behave depends on their type:
- Mails and AI tokens (cheap, own key from Pro onwards): soft limit at 110%, hard limit at 120%, a buffer for occasional spikes
- Database and file storage (expensive and persistent): soft limit at 100%, hard limit at 105%, stricter because the data sits on our disks
- Number of apps (a slot on the platform): hard limit at 100%, no buffer, because it is booked capacity
- Outbound traffic: soft limit at 110%, hard limit at 130%, because traffic spikes are normal
- When the hard limit is reached, no more data can be written (error code HTTP 429). Reading stays possible, and the way to the bigger plan is shown immediately
- At disproportionately high load (more than 10 times the plan median over 7 days) we talk to you. We don’t suspend anything by surprise
- Caps reset every calendar month, on the 1st at 00:00 UTC. AI tokens have no reset, the 7-day window slides along
Not capped:
- Tenants inside your app (internal tenants are yours, any number)
- Workflow complexity, number of data objects and volume of events inside your domain logic
- Your app logic, as long as it doesn’t blow our platform resources
Custom solutions on request
The standard plans cover 95% of what DACH mid-market apps need. If a plan doesn’t fit, we don’t push you up, we talk to you.
| What you might need | What we do |
|---|---|
| More performance (bigger app, more compute, more memory, faster database) | a custom performance class on top of your existing plan, no jump to Enterprise required |
| Higher availability (for example 99.9% instead of standard) | a custom SLA (agreed availability), scoped to what your business needs |
| Encryption with your own key (your vault, your keys) | available when your compliance rules out keys managed by Kumiko |
| Dedicated stack (own database cluster, own namespace) | for isolated performance or strictly regulated industries |
| A higher cap for a single value that doesn’t fit | we adjust that one cap, not the whole plan |
The plan gives you a clean default, and when it doesn’t fit, a custom solution goes on top. There are no surprise surcharges for being 3% over. We talk first, agree on the scope, and then start.
Pilot program (today)
Mid-market companies from the DACH region in the pilot program get 3 months free, plus a joint setup phase of about 3 weeks. In return we may write a case study about you, and you can veto any sentence. Just ask whether a place is currently open. Requirements are 50 to 500 employees, your own IT and a concrete first use case.
Details: [email protected]
Open questions until Q1 2027
- The entry price for Pro is a guide value of €49. It is open upward, depending on real per-app hosting costs and pilot feedback
- The definition of the Business plan: where is the line between “I need multiple apps” and “I need an enterprise contract”?
- Pricing for custom performance classes: what is the right price list for bookings with special performance or a special SLA?
- The demo cap on the Free plan: how many AI tokens are enough to try things out without costs running away?
Answers come once the first pilot conversations become concrete.
License background
BSL-1.1 with conversion to Apache-2.0 after 2 years, the same model Sourcegraph uses. Free to use for everything except commercial hosting that competes with kumiko.rocks. This protects against third parties reselling it as a cloud service, and the open-source story stays complete. Decided on 2 May 2026.