Each request supplies or identifies the information needed to process it, allowing different compatible server instances to handle successive calls. This can simplify distribution and scaling compared with a design tied to one persistent client-server session.
Stateless does not mean the application stores no data. Durable records, authorization and other state can still exist, but the request path must manage them without assuming an earlier conversation remains in one server's memory.
ELI5
Think of sending a complete service request that another worker can understand without having handled your previous request.
For example, a tool request can carry the required operation and identifiers so an available server can process it without a dedicated ongoing conversation.
Can a stateless service use a database?
Yes. It can access durable application data without depending on a particular server retaining the client's conversational session.
Does stateless design remove compatibility checks?
No. Clients and servers still need compatible formats, supported operations and authorization.

