A cloud platform for building and running enterprise applications.
Applications authored as metadata — forms, screens, REST endpoints and rules materialised live without redeploys. One supervised runtime across seven OLTP engines and three OLAP engines, no separate BI stack. AI built in — multi-provider, MCP-native, governed by the same security perimeter as every other workload. The enterprise ERP, CRM, HR and project-management suite in the Solutions section runs on this runtime today.
Metadata-driven runtime
Applications are metadata, not artefacts. Endpoints, screens, triggers and roles are SQL rows the runtime materialises live. Hot-reload propagates cluster-wide in milliseconds via Redis pub/sub. No JARs, no redeploys.
OLTP and OLAP — addressed natively
Native SQL generation for every supported engine — no generic dialect to translate at runtime. OLTP across PostgreSQL, Informix, Oracle, DB2, SQL Server, MySQL and SAP HANA; OLAP across Vertica, ClickHouse and BigQuery. The application picks the engine its workload deserves; the compiler emits engine-native SQL either way.
AI-native, MCP-native
Multi-provider abstraction across OpenAI, Anthropic, Ollama, Google Vertex, IBM Watson and Cohere. Built-in RAG over the customer's choice of vector store — Qdrant, Milvus, PostgreSQL pgvector or Redis. Native MCP server — AI assistants operate inside the user's permission perimeter, never above it.
One cockpit for the entire estate
Studio centralises every runtime concern — servers, databases, users, roles, certificates, AI providers, cron, integrations and audit — into one governed surface. Eight authentication mechanisms under one dispatcher. Observability as SQL : SELECT * FROM jvm.memory replaces the APM bill.
The platform's surface area, in numbers.
Every claim above is enumerable. The counts are the spec.
How most enterprise platforms ship — and how Airtool ships
| Traditional application server | Generic visual app builder | Airtool | |
|---|---|---|---|
| Application model | Compiled artefacts (WAR/JAR), redeploy required | Visual editor, vendor-locked output | ✓ Metadata in a Metadata Repository, materialised at runtime, hot-reloaded |
| Cluster cache invalidation | Manual or never | Vendor-managed | ✓ Redis pub/sub, milliseconds across nodes |
| Database support | Single engine, locked at architecture | Vendor-supported list, often abstracted | ✓ Seven OLTP engines for the application; three analytical engines for OLAP — all on the same source |
| AI integration | External vendor, separate perimeter | Vendor's chosen LLM, governance opaque | ✓ Multi-provider, governed, MCP-native, cost-tracked |
| Observability | External APM with separate licence | Vendor portal | ✓ In-memory MemDB, queryable as SQL — no APM bill |
| Security model | Bolted on per application | Vendor-managed, unauditable | ✓ 8 unified auth mechanisms; row, column and SQL injection by default |
Airtool Apps — included or extended
An enterprise application platform with a business suite in production at customer sites today — ERP, CRM, HR and project management — and bespoke customer applications engineered on the same runtime. Multi-tenant by construction. Non-stop life-cycle. Source-available.
The applications are the platform's engineering proof — multi-tenant, non-stop, SaaS-capable by construction.
A platform's most credible claim is what it already runs. The Airtool runtime runs a business suite — ERP, CRM, HR and project management — in production at customer sites today, alongside the bespoke applications customers engineer on the same runtime. Engineering depth is demonstrated, not described.
The runtime is multi-tenant by construction. One installation hosts many companies — each with its own data, security perimeter and audit trail — with tenant boundaries enforced at the JDBC layer rather than in application code. No Kubernetes orchestration layer. No service-mesh sidecar. No separate scaling control plane to operate. Nodes are stateless ; cache invalidation propagates cluster-wide in milliseconds through Redis pub/sub. Adding capacity is adding a node. The application life-cycle is non-stop : new metadata, new screens and new logic propagate without a deployment window.
The structural difference is who owns the engineering platform. SAP, Salesforce and Oracle each run on their own. What they sell are the applications and the extension surfaces that orbit them — BTP, Lightning, OCI — while the engineering platform itself stays inside the vendor. Airtool inverts the model : the runtime that runs the Airtool business suite is the runtime customers run their own enterprise applications on. The proof is production today, not a roadmap commitment.