Apps and Platform. SaaS or on-prem.

Because tiered plans hide pricing rather than communicate it. The platform's capabilities — supervised execution, multi-tenancy, audit, observability, AI inside the perimeter — are not features to be unlocked at higher tiers ; they are properties of the runtime that every engagement inherits. One rate per surface, with volume and term-length adjustments scoped in the engagement rather than in a marketing table, is the honest expression of how the business is actually priced.

The Platform licence is what teams need when they author their own applications on Airtool — the Application Editor, the standard library, the AI development agents, the administration surface, the full build / administer / operate toolchain. The Apps licence covers running the packaged ERP / CRM / HR / PM applications. Customers who license Apps and add Platform later license both. Customers who license Platform without Apps are building their own application surface on the runtime.

Production runs on a 2-node high-availability cluster, minimum. Each non-production environment — DEV, QA, TEST, staging — typically runs on a single node, and multiple non-production environments often share one node because the runtime is multi-tenant by design and the environment context isolates them. A typical small mid-market footprint is 2 production nodes plus 1 shared non-production node (3 nodes total) ; a typical large mid-market footprint adds a dedicated DEV node and a batch / headless worker node (4 — 5 nodes total) ; an enterprise deployment scales the cluster horizontally by booting more application nodes (6, 8, more — same model).

Adding capacity is a node boot : each node reads its role from configuration (production application, non-production environment, batch worker, AI orchestrator), registers with the control center at its boot URL and auto-discovers its peers. No Kubernetes operator estate to author, no Helm charts to maintain, no service mesh to configure — the platform is a single supervised runtime, so adding capacity is a node boot and a role assignment, not a deployment programme.

Airtool is an enterprise application platform — one supervised runtime that authors applications as metadata, compiles native SQL across seven OLTP engines and three OLAP engines, runs AI inside the perimeter, ships audit-grade discipline by default, and operates from a 2-node high-availability cluster with role-based node boot. The Apps suite — ERP, CRM, HR, PM — is source-available and serves as a working reference architecture customers can read, fork and extend.

Airtool is not a visual workflow tool for departmental apps ; not a forms-over-data layer ; not a relational spreadsheet ; not a microservice framework that needs a Kubernetes orchestration estate underneath it ; not a vendor-bundled foundation tied to a single database. There is no clean one-to-one comparable in the market because that combination is uncommon. The question we suggest a buyer ask is not which tool is closest to Airtool but what foundation their applications will run on in five and ten years' time, and on what assumptions about ownership, audit, AI, database choice and operational complexity.

Comparing Airtool to those tools is comparing an airplane to a bicycle — both are vehicles, but they answer different questions and sit in different categories. Mendix, OutSystems and Power Apps are visual application builders sized for the next departmental form, workflow or dashboard ; Airtable is a relational spreadsheet with a UI on top. Airtool is an enterprise application operating system — one supervised runtime that authors applications as metadata, compiles native SQL across seven OLTP engines and three OLAP engines, runs AI inside the perimeter, ships audit-grade discipline by default, and runs the strategic application estate of an enterprise for a decade or more. A feature-by-feature comparison treats them as substitutes, which they are not. Request a demo — twenty minutes side-by-side on your own data model, with a commercial architect walking through it, makes the category difference visible before the feature lists do.

No. Platform deployments require at least two developers. A platform with fewer than two developers and a viable user fleet is not a real engagement — it is a single point of failure with no review surface. Application-only engagements that do not need a platform licence are served by the Apps tiers.

No. The AI provider's API tokens are billed by the provider you choose ; the platform's role is the abstraction, governance and cost-dashboard layer on top, which is included. The customer chooses the provider — and the provider's commercial relationship — and the platform makes the choice configurable rather than locking it.

SaaS is hosted by the platform team on AWS infrastructure operated by the platform team — predictable, no infrastructure for the customer to run. Apps on SaaS are priced per user, with rates that scale by value-per-seat to the function — from $6 / user / month for HR, $20 for PM, $30 for CRM, $99 for ERP and $119 for the All-Apps suite (23% below the sum-of-modules). Platform on SaaS is scoped per engagement in the commercial conversation, with AWS Marketplace AMI as the planned self-serve path. On-prem mirrors the SaaS Platform structure on customer infrastructure at a rate below the equivalent SaaS line, with survival rights on the deployed runtime — if the subscription lapses, the deployed runtime keeps running on customer infrastructure indefinitely. Implementation services are scoped separately in every case and quoted phase by phase.

AWS Marketplace AMI is coming soon — the self-serve path on AWS infrastructure, launched on the customer's AWS account in minutes, ideal for proof of concept and small production deployments. It will be the primary public path for Platform SaaS when it launches ; in the interim, the engagement path is the commercial architect conversation. AI provider API tokens (OpenAI, Anthropic, Vertex, Watson, Cohere, Ollama) are billed by the provider at provider list, with zero markup, regardless of the deployment path.

On-prem is sold as an annual subscription, and the subscription buys software updates, AI features, support and monitoring. If the subscription is not renewed, the deployed runtime continues to run on customer infrastructure indefinitely — there is no licence-check phone-home, no time-bomb, no remote kill-switch on the application that has been deployed. The applications your team has built keep serving users ; the integrations keep firing ; the batch jobs keep running ; the data stays on the customer's storage. What stops is the flow of new versions, the AI services that depend on the platform team's cloud, the support relationship and the monitoring telemetry. This is the licence model HashiCorp, GitLab and Confluent ship their enterprise platforms under, and the shape procurement teams expect on strategic infrastructure that the business depends on for a decade or more.

A fit conversation, not a checkout flow. A response within two business days from someone qualified to write the trade-offs down. Discovery call within 48 hours of the first conversation, architecture review within four weeks, pilot within one quarter. The licence surface is the pricing on this page ; the implementation surface is the engagement, scoped and quoted in the same conversation.