As per MongoDB documentation here The Ops Manager Backup Agent provides scheduled snapshots and point-in-time recovery of your MongoDB replica sets and sharded clusters.
How it Works
A lightweight Backup Agent runs within your infrastructure and backs up data from the MongoDB processes you have specified.
NOTE : Only sharded clusters or replica sets can be backed up. To back up a standalone mongod process, you must first convert
it to a single-member replica set.
When you start backing up a MongoDB deployment, the agent performs an initial sync of the deployment's data as if it were creating a new,
invisible member of a replica set. For a sharded cluster, the agent syncs each shard’s primary member and each config server. The agent sends the initial sync and oplog data over HTTPS back to Ops Manager.
The Backup Agent then tails each replica set's oplog to maintain on disk a standalone database, called a head database. Ops Manager maintains one head database for each backed-up replica set. The head database is consistent with the original primary up to the last oplog supplied by the agent.
The Backup Agent executes the initial sync and the tailing of the oplog using standard MongoDB queries. The cluster being backed up is unaware of the additional copy of the backup data.
The Backup Agent uses a MongoDB instance version equal to or greater than the version of the replica set it backs up.
The Backup Agent takes and stores snapshots based on a user-defined snapshot retention policy. Sharded cluster snapshots temporarily stop the balancer so that they can insert a marker token into all shards and config servers in the cluster. Ops Manager takes a snapshot when the marker tokens appear in the snapshot data.
For further your ref here , here and here