I believe it's a low-level "toolbox" to allow xml-based "database" storage on a local computer. It does include a command line shell that's useful for exploring the capabilities.
To do anything useful, you need to write a front end, using the low-level API's that sleepycat provides.
Note that this database is primarily designed for single-computer use. If you want to extend it for multiple, concurrent users, you would want to write a service (probably using TCP/IP) to act as a middle layer between the data files and your clients, otherwise, you would run into the same file system based concurrency issues of which MS Access suffers.
As does the non-xml version of sleepcat, it creates a separate folder and multiple files within the folder on the target computer for each "table" it creates. Since it uses a native XML format for storing the data, it is perhaps more suitable for semi-structured data than is the "standard" version of sleepycat.
all the API's are called in-process, so that it can be more responsive than accessing a client-server based database, which depends on inter-process communication between the client and the server.