The Kirak runtime is open source and free forever. Kirak Studio is the managed service on top — the AI build agent, dedicated hosting, and the operating console. Pick a plan by how many projects you run and how much compute they need. A Studio project checks out to your own GitHub and runs the identical runtime, so choosing Studio now is not a one-way door.
Build and deploy a single project on Studio.
For side projects and prototypes that need to stay up.
For teams shipping production apps.
For studios running many client projects.
The full runtime, every module included — no account, no key, no phone-home. Host it on any infrastructure you like.
Prices are placeholders — final Studio and instance pricing is being finalized.
Every project runs on one or more dedicated instances. Start on a free Staging instance, scale up when you go live. Billed hourly, capped at the monthly rate — stop an instance and billing stops.
| Instance | vCPU | RAM | Hourly | Monthly cap |
|---|---|---|---|---|
| Staging | shared | — | Free | Free — 3-day limit, for testing |
| Small | 2 | 4 GB | $—/hr | $19/mo |
| Medium | 4 | 8 GB | $—/hr | $39/mo |
| Large | 8 | 16 GB | $—/hr | $74/mo |
| Horizontal scaling — HA | multi-node | — | — | Coming soon |
Hourly rates and final monthly pricing are placeholders. Larger instances and custom sizing are available on Agency and Enterprise.
Kirak's Small instance is a fully isolated 2 vCPU / 4 GB VPS — single-tenant, full CPU, root access. Here's the nearest tier on the API-backend services (Supabase, Xano) and the modern hosting platforms (Render, Railway) people weigh against it.
| Comparable tier | vCPU / RAM | Compute price | Base / platform fee | Effective / mo | CPU type |
|---|---|---|---|---|---|
| Kirak · Small | 2 / 4 GB | ~$19 | $0 | ~$19 | Dedicated VPS, single-tenant |
| Supabase · Medium compute | 2-core / 4 GB | $60 | Pro $25 incl. $10 compute credit | ~$75 | Shared dedicated from 8 GB up |
| Xano · Essential | Not published ≈1 core; more via paid boost | $85 billed annually | $0 | $85+ | Dedicated instance spec undisclosed |
| Render · 2c-4g | 2 / 4 GB | $85 | Workspace $0–$25 | $85–$110 | Allocated, managed container |
| Railway · 2 vCPU / 4 GB | 2 / 4 GB | $80 2 × $20 vCPU + 4 × $10 GB, 24/7 | Hobby $5 / Pro $20 credited against usage | ~$80 | Shared, usage-metered |
| Appwrite · 4 GB / 2 CPU runtime | 2 / 4 GB | ~$436 metered by GB-hour, not a fixed instance | Pro $25 | ~$461 | Serverless functions no persistent instance |
Public pricing, September 2026 — verify current rates before relying on these; Kirak prices are placeholders. Supabase dedicated CPU starts at its 8 GB tier ($110+). Xano bills annually and doesn’t publish per-plan CPU/RAM; Essential is roughly one dedicated core, with more sold as a paid boost. Render allocates the stated vCPU but runs a managed multi-tenant platform, not a VPS. Railway bills by the minute — $20 per vCPU and $10 per GB of RAM per month — so the figure assumes a service running continuously; one that idles or scales to zero costs less. Appwrite meters function runtime by GB-hour: the figure assumes its 4 GB / 2 CPU spec (8 GB-hours per running hour) serving traffic all month — 8 × 730 h = 5,840 GB-hours, less the 1,000 included on Pro, at $0.09/GB-hour ≈ $436 — plus the $25 Pro fee; near-idle it costs far less.
Plans differ only in how many projects and instances you run and how AI is billed. Every Studio console feature — build agent, model editor, API tools, deploys, data, monitoring, GitHub sync — is on every tier.
| Free | Hobby | Pro | Agency | |
|---|---|---|---|---|
| Plans & limits | ||||
| Projects | 1 | 3 | 10 | Unlimited |
| Dedicated instances included | Staging only | 1 | 3 | BYOC |
| Extra instances | — | Pay-as-you-go | Pay-as-you-go | Your cloud |
| Team seats | 1 | 1 | Team | Team |
| AI build agent usage | Included | Included | Extra credits | BYOK |
| Bring your own cloud (BYOC) | — | — | — | ✓ |
| Support | Community | Priority | Priority | |
| AI build describe it, Kirak generates it | ||||
| Conversational spec-building — live readiness matrix, safe defaults | ✓ | ✓ | ✓ | ✓ |
| Starter templates — Stripe API, MCP agent, mobile auth | ✓ | ✓ | ✓ | ✓ |
| Project overview map + next-steps checklist | ✓ | ✓ | ✓ | ✓ |
| Kirak.md spec export for other AI tools | ✓ | ✓ | ✓ | ✓ |
| Models & logic | ||||
| Visual model editor + ERD view | ✓ | ✓ | ✓ | ✓ |
| Row- & column-level security, per-model rate limits | ✓ | ✓ | ✓ | ✓ |
| Staged edits with migration safety warnings | ✓ | ✓ | ✓ | ✓ |
| Lifecycle hooks — visual Logic Flow + Python editor, kept in sync | ✓ | ✓ | ✓ | ✓ |
| Hook AI Copilot, Test Bench, Dry Run | ✓ | ✓ | ✓ | ✓ |
| Custom API endpoints | ✓ | ✓ | ✓ | ✓ |
| API & docs | ||||
| Auto-generated REST API, regenerated as models change | ✓ | ✓ | ✓ | ✓ |
| GraphQL API — per-model opt-in | ✓ | ✓ | ✓ | ✓ |
| OpenAPI 3 + Markdown docs export | ✓ | ✓ | ✓ | ✓ |
| Interactive playground — mock data, multi-environment | ✓ | ✓ | ✓ | ✓ |
| curl / JS / Python snippets per route | ✓ | ✓ | ✓ | ✓ |
| MCP tools exposed as callable endpoints | ✓ | ✓ | ✓ | ✓ |
| Deploy & operate | ||||
| One-click deploy, redeploy, roll back — version tracked | ✓ | ✓ | ✓ | ✓ |
| Vertical resize (S/M/L/XL), zero-downtime reload | ✓ | ✓ | ✓ | ✓ |
| Multiple environments per project — prod, staging, per-client | ✓ | ✓ | ✓ | ✓ |
| Clone instance | ✓ | ✓ | ✓ | ✓ |
| Encrypted secret store — sealed per environment, reveal / rotate | ✓ | ✓ | ✓ | ✓ |
| Guided env setup — per-service keys, test connection | ✓ | ✓ | ✓ | ✓ |
| Custom domains + managed SSL | ✓ | ✓ | ✓ | ✓ |
| Data & monitoring | ||||
| Data browser — tables, record editing, universal search | ✓ | ✓ | ✓ | ✓ |
| Dashboard widgets + time-series charts | ✓ | ✓ | ✓ | ✓ |
| Saved views / multi-table joins + CSV export | ✓ | ✓ | ✓ | ✓ |
| Live logs + distributed traces | ✓ | ✓ | ✓ | ✓ |
| Alert rules — error rate, p95/p99, uptime, cron failure | ✓ | ✓ | ✓ | ✓ |
| Notification channels — Email, Slack, Webhook, PagerDuty | ✓ | ✓ | ✓ | ✓ |
| Scheduled jobs — run now, run history, on-failure notify | ✓ | ✓ | ✓ | ✓ |
| Source & collaboration | ||||
| Two-way GitHub sync — change list, per-change undo, diffs | ✓ | ✓ | ✓ | ✓ |
| Commit & push to main, or open a Pull Request | ✓ | ✓ | ✓ | ✓ |
| Global search across projects, models, hooks (⌘K) | ✓ | ✓ | ✓ | ✓ |
| Team management — invite teammates | 1 seat | 1 seat | ✓ | ✓ |
| Everything file-backed and versioned — models, hooks, docs | ✓ | ✓ | ✓ | ✓ |
Some Studio console items are target design and still being wired up — instance autoscaling in particular. Prices are placeholders.
You pay for projects, compute, and operations — never for capabilities. Every Studio plan and every self-hosted install serves the full runtime.
AGENTS.md, JSON Schemas, kirak validate, and a dev MCP server in every project.Database-agnostic — MySQL or PostgreSQL, one config line · SOC 2–ready security controls · no lock-in, Apache 2.0 the whole way down.
Beyond Agency: custom project and instance limits, dedicated support with an SLA, SSO/SAML, private-cloud or on-prem deployment, and help with procurement, security review, and compliance.
A project is one deployed backend — its own models, kirak.json, hooks, database, and instances. Staging and production environments of the same project don't count separately.
A dedicated VPS that runs your project's runtime. Every project needs at least one; you add more to scale or to separate environments. Instances are billed by the hour, capped at the monthly rate.
Shared platforms give you a fraction of a CPU that throttles once you burn through burst credits, and your neighbors' traffic affects your latency. A dedicated instance reserves its full CPU and RAM for you, so performance stays flat under load — you can run more on a smaller, cheaper box.
Hourly, based on the instance size, capped at the monthly rate shown above. Stop an instance and billing stops. Staging instances are free for their 3-day window.
Yes, up or down at any time. Upgrades take effect immediately; downgrades at the end of the billing period. If you hit a limit, we prompt you to upgrade rather than cutting anything off.
BYOC (bring your own cloud) runs your instances in your own AWS/GCP account — Kirak Studio operates them, but the infrastructure bill is yours. BYOK (bring your own key) points the AI build agent at your own OpenAI, Anthropic, or Gemini key instead of Studio's included credits.
Any time. A Studio project checks out to your own GitHub — models, kirak.json, hooks, custom endpoints — and runs the identical open-source runtime. No rewrite, no lock-in.
The Free plan is free forever for one project. Start there and upgrade when you're ready — there's nothing to cancel if you stay on Free.