Your options.
Side by side.

There is no universally best data framework. Storm is built for teams, and the coding agents working alongside them, who want explicit, predictable database access with concise, immutable models. Here is where it sits next to the alternatives, and where each of them is the better call.

At a glance

Decision-relevant differences across the most common choices. This page compares the designs; the measured numbers, on identical workloads, live on the benchmarks page.

Feature	Storm	JPA / Hibernate	jOOQ	Exposed	Ktorm
Entity model	Immutable data class (~5 lines)	Mutable class (~30, ~10 with Lombok)	Generated from schema	DSL table object (+ optional DAO)	Mutable interface (+ DSL table)
Immutable entities	Yes	No	Yes	DSL only	No
Type-safe queries	Yes	Criteria API	Yes	Yes	Yes
N+1 handling	Single-query entity graph	Common pitfall	Manual	Manual joins (DSL) · eager-load (DAO)	Manual
Query across relations	One line: joins and mapping derived from @FK	JPQL strings or Criteria builders	Implicit path joins + multiset	Manual joins and row mapping	Manual joins
Session state	None: no session, no flush	Persistence context, flush, dirty checking	None	DSL none · DAO transaction-bound	Per-entity change tracking
Deferred loading	Explicit Ref<T>	Proxies, session-bound	Not applicable	DAO lazy, transaction-bound	Manual joins
Standalone transactions	Propagation, isolation, timeout, commit hooks	EntityTransaction · JTA/Spring for more	Lambda API, nested	Nested, isolation config	Lambda API
Row mapping	Compile-time generated, reflection-free	Reflection / bytecode	Generated records	Manual (DSL)	Dynamic proxies
GraalVM native images	Native support for Spring and Ktor	Via Spring Boot AOT or Quarkus	Reflection config for generated records	Via the Spring Boot integration	Manual configuration
Full SQL escape hatch	SQL / SQL templates	Native queries	It is SQL	Raw exec()	Raw JDBC
Languages	Kotlin + Java	Kotlin + Java	Kotlin + Java	Kotlin only	Kotlin only
License	Apache 2.0	LGPL 2.1	Commercial for some DBs	Apache 2.0	Apache 2.0

When an agent writes the code

Three rows above decide how well a coding agent does with a framework. Entity model: in Storm one data class is the table, its keys and its relations, so the agent reads one file and has the full picture. Session state: with none, the code the agent reads is exactly what runs, and no flush or proxy acts on it later. Query across relations: the domain model already gives one-line, type-safe queries with the whole entity graph loaded, so there is less for the agent to write and fewer joins and mappings to get wrong.

On top of that, one command, npx @storm-orm/cli init, installs Storm's rules and skills for Claude Code, Cursor, GitHub Copilot, Windsurf and Codex, plus a local, read-only MCP server that gives the agent your schema without your credentials. See the agentic coding guide.