Status: Working draft. Final pricing will be set Q3 2026 with the first pilot customers.
What’s already settled
- Framework + bundled features are public — BSL-1.1 on GitHub (
kumiko-framework/kumiko), NPM packages free to use. Every release automatically becomes Apache-2.0 after 2 years - Platform IP (Designer UI, AI-builder, tenant provisioning, billing) stays in the private repo — protection against cloud clones
- kumiko.so Hosted is tier-based
- BYOK is the default, not a premium feature — your variable costs belong in your accounts (see below)
- No per-user pricing — unlike Retool. At 100 employees you don’t pay 100× more
Tier sketch
| Tier | Price range | Apps | Hosting | AI | Support | Audience |
|---|---|---|---|---|---|---|
| Self-Host (Open Source) | €0 license, ~€50–150/month server | n | Your choice | BYOK | Community | Self-host builders, mid-market with IT |
| Free (Hosted) | €0 | 1 | kumiko.so/sub | Anthropic demo cap | Community | Indie hackers, MVPs |
| Pro (Hosted) | ~€49–149/month per app | 5 | kumiko.so/sub or own domain | BYOK | Solo devs, small teams | |
| Business (Hosted) | TBD (~€300–500/month) | 20 | Like Pro + raised caps | BYOK | Email + priority | Growing teams, multiple apps |
| Enterprise | Mid 4–5 figures p.a. | per contract | Dedicated or on-prem | Local model or BYOK | SLA + premium | European mid-market, regulated industries |
Pro / Business range is per app, not per user. Apps with multiple tenants run on a single instance.
Tier upgrades without data loss
Moving up a tier doesn’t migrate your data — it doesn’t touch your data. We change the resources your app runs on, the database stays exactly where it was. The whole upgrade is on the order of seconds of restart, no export, no re-import, no “downtime window” that turns into a downtime weekend.
The same is true going down. If you over-bought, scale back without losing anything. The data outlives the tier choice.
Why this matters in the pitch: Most internal-tools platforms make tier changes a “talk to sales” event because the migration is painful for them. We made it boring on purpose — the boring path is also the safe path.
Your variable costs belong in your accounts
What’s variable runs through customer accounts — not ours.
No platform-default key as a business model. Three consequences:
- Your reputation, your control. When you send email to your customers, that runs through your Brevo / Postmark account, not ours. Spam issues and reputation damage hit you — not 50 other tenants on the platform
- Your real costs are visible. Anthropic bill at Anthropic, S3 bill at Hetzner, SMTP bill at Brevo. The platform tier covers platform operations, not your externals
- Compliance bonus. “Your data, your AI, your mail never leave your accounts” — same architecture lever as for European on-prem buyers
| Service | Default from Pro onwards | Free onboarding |
|---|---|---|
| AI tokens | BYOK required (Anthropic / OpenAI / Ollama) | 100k-token demo cap with platform key |
| Email send | BYOK required (Brevo / Postmark / SES SMTP) | 1,000 platform mails/month (login, reset, notifications) — no mass mailing |
| File storage | BYO-S3 (Hetzner Object Storage / MinIO / AWS-S3) | 1–10 GB platform cap, then BYO |
Caps per tier
We cap platform resources, not your business logic. What your app does internally — tenants, workflows, data structure — is yours. We only measure what flows into our Postgres, our mail server and our GPU pool.
| Cap | Free | Pro | Business | Enterprise |
|---|---|---|---|---|
| Apps | 1 | 5 | 20 | per contract |
| DB storage (summed across apps + internal tenants) | 100 MB | 10 GB | 100 GB | per contract |
| File storage (platform) | 1 GB | 10 GB / BYO-S3 | 100 GB / BYO-S3 | BYO-S3 |
| Platform mails/month (login/reset) | 1,000 | 10,000 / BYO-SMTP | BYO-SMTP required | BYO-SMTP |
| AI tokens (rolling 7 days) | 0 (no AI Designer on Free) | 50,000 | 150,000 | BYOK / local |
| Egress (bandwidth) | 10 GB/month | 100 GB/month | 500 GB/month | per contract |
Cap behavior — staggered by cap type:
- Mails / AI tokens (cheap, BYOK from Pro onwards): soft @ 110 %, hard @ 120 % — burst buffer for sporadic use
- DB storage / file storage (expensive + persistent): soft @ 100 %, hard @ 105 % — stricter, ends up on our disk
- Apps count (platform slot): hard @ 100 % — no buffer, it’s booked capacity
- Egress: soft @ 110 %, hard @ 130 % — bursty traffic is legitimate
- Hard cap reached: write operations blocked (HTTP 429), reads stay open, upgrade path immediately visible
- Fair use: at disproportionate load (>10× tier median over 7 days) we talk to you — no surprise suspensions
- Reset: calendar month, 1st at 00:00 UTC. AI tokens use a rolling 7-day window (no reset, slides)
What we don’t cap:
- Tenants inside your app (internal multi-tenancy is yours, any number)
- Workflow complexity / aggregate count / event volume inside your domain
- Your app logic — as long as it doesn’t blow our platform resources
Custom on request
Standard tiers cover 95% of what DACH mid-market apps need. When the tier doesn’t fit, we don’t force-upgrade you — we talk.
| What you might need | What we do |
|---|---|
| More performance (bigger app, more CPU, more RAM, faster DB) | Custom performance class on top of your existing tier — no jump to Enterprise required |
| Higher availability (e.g. 99.9% instead of standard) | Custom SLA, scoped to what your business actually needs |
| BYOK encryption (your vault, your keys) | Available when your compliance requires no Kumiko-managed keys |
| Dedicated stack (own DB cluster, own namespace) | For performance isolation or strictly regulated industries |
| Raised caps for a single dimension that doesn’t fit | We adjust the one cap, not the whole tier |
The principle: tier defines a clean default, custom on top when the default doesn’t fit. No surprise auto-charges for “you ran 3% over”. We talk first, scope it, then run it.
Pilot program (today)
3–5 European mid-market customers, 3 months free, joint ~3-week setup sprint. In return: reference-story right (with veto on every sentence). Threshold: 50–500 employees, own IT, a concrete first use case.
→ Details: hello@kumiko.so
Open questions for Q3 2026
- Pro tier price within €49–149 — depends on real per-app hosting cost and pilot feedback
- Business tier definition — where’s the line between “needs multiple apps” and “needs an enterprise contract”?
- Custom-class pricing — what’s the right rate card for non-standard performance/SLA bookings?
- Free tier demo cap — how much AI-token demo is enough to try without cost explosion?
Answers come once the first pilot conversations are concrete.
License background
BSL-1.1 with 2-year conversion to Apache-2.0 — the Sourcegraph model. Free to use for everything except commercial hosting that competes with kumiko.so. Protection against cloud re-sale + full open-source story. Decided 2026-05-02.