Iron Lake is a small studio building infrastructure for the agents coming next. Owned hardware. Published pricing. Permanent free tier. No web console. No sales team. No hidden fees.
▸ Our first product, Kingdom, ships Q2 2026.
And commodities should be priced like commodities. We publish our pricing formula. We cap gross margin at 30–40%. If a CPU-second costs X to produce, we charge Y where the ratio is public and versioned. No surprise per-API-call fees. No egress gouging at $0.09 per GB when the underlying transit costs 40× less. Your bill is compute, storage, and bandwidth — nothing else.
Renting from hyperscalers caps our cost floor at theirs. We run on hardware we bought, in a rack we pay for. This makes us slower to scale and much cheaper to run. We're fine with the trade. When you sit on metal you own, price pressure flows one direction only.
Every cloud today was designed for humans clicking through consoles. Agents use them by imitation — screen-scraping dashboards, driving browser automation harnesses, pretending to be the human the interface expects. We invert this. The API is the product. A skill file teaches agents to use it. Dashboards, if they exist, are read-only reflections of API state. You don't need to be a person to deploy here. You need to be a tenant in good standing with a valid API key.
Not a trial. Not credits. A floor. Autonomous agents need a place to try things without a credit card in the prompt, and the cost of running a baseline tenant is smaller than the cost of acquiring one. We'll be fine.
No enterprise contracts. No negotiated SKUs. No sales team to fund with your bill. Every feature is available to every account. The pricing is the pricing. The interface is the interface. If that model can't sustain us, we don't deserve the business.
A home for agents.
An agent reads a skill file. It makes four HTTP calls. Its workload
boots in a Firecracker microVM on our metal, receives a *.kingdom.sh URL, and is reachable on the public internet
over HTTPS within thirty seconds of submission. Slurm schedules. Rust
serves. Cost metadata rides on every response. Budget enforcement
lives in the scheduler, not the API, because the race condition
matters.
$ curl -X POST https://api.kingdom.sh/v1/jobs \
-H "Authorization: Bearer $KINGDOM_KEY" \
-d '{"image": "ghcr.io/you/yourapp:latest",
"ingress_port": 8080}'
{
"id": "kj_01HW8Z3N5K...",
"state": "BOOTING",
"public_url": "https://hello-a4b9.kingdom.sh",
"cost": {
"this_request_usd": 0.00,
"period_total_usd": 0.00,
"budget_remaining_usd": null
}
}Footnote. The demo target is a ≤2-minute recording of the above, filmed on homelab hardware visible in-frame. The homelab platform (multi-tenant, VXLAN-isolated, metered, safety-circuited) follows in Q3. Migrations of our own two production products — a marketing-SaaS for medspas, and a fan-in form builder — into Kingdom are Q4.
For the people who ask "but what is it actually made of."
A deeper technical manifesto — the research that shaped these choices, 20 deviations from our own original spec, the five catastrophic failure modes we designed against before writing a line of code — is coming when Kingdom opens.
We'll email you once when Kingdom opens the first wave of accounts. We won't email you for any other reason. We don't sell your address or enrich it. If we ever have to, we'll delete it and apologize.