You could do this using a scheduled script(cron / windows scheduled task) that communicates to SplunkD via the Splunk REST API.
The script could execute a search(via the REST API) to determine the last time each user logged in, and calculate which users have been inactive for x days.In the following example, 2592000 seconds is 30 days
index=_internal source=*web_service.log action=login status=success | eval last_login_time=_time | eval current_time=now() | eval time_since_last_login_secs=current_time-last_login_time | where time_since_last_login_secs > 2592000 | table user
And then remove the users(via the REST API) that are returned in the search results.
We have several language SDKs that sit atop the REST API to make this development easier.
You could also do this using the Splunk CLI.