1) In a stage/pipeline, use an OutputStage?. The stage policy should look like:

OutputItems: {
    {kernelKey}: {
        StoragePolicy: {
            Storage: "XmlStorage"
            Location: "%(output){relative path}"
        }
    }
}

2) In a test script, use an OutputStage? created with SimpleStageTester.

from lsst.pex.harness import SimpleStageTester
from lsst.pex.policy import Policy
pol = Policy()
pol.set("OutputItems.d.StoragePolicy.Storage", "XmlStorage")
pol.set("OutputItems.d.StoragePolicy.Location", path)
stage = SimpleStageTester.create("lsst.pex.harness.IOStage.OutputStage", pol)
clip = stage.run({"d": object}, 0)

3) In a test script, you could also use lsst.daf.persistence directly:

from lsst.daf.base import PropertySet
import lsst.daf.persistence as dafPersist
from lsst.pex.policy import Policy
pol = Policy()
persistence = dafPersist.Persistence.getPersistence(pol)
storageList = dafPersist.StorageList()
loc = dafPersist.LogicalLocation(path)
storage = persistence.getPersistStorage("XmlStorage", loc)
storageList.append(storage)
persistence.persist(object, storageList, PropertySet())

To recover this thing from disk for e.g. a pcaPsf

becker127: more psfModel.xml 
...
<psf class_id="0" class_name="lsst::meas::algorithms::pcaPsf" tracking_level="1" version="0" object_id="_0">
...

I need to say

import sys
import lsst.daf.persistence as dafPersist
import lsst.daf.base        as dafBase
import lsst.pex.policy      as pexPolicy
import lsst.meas.algorithms as algorithms

import pdb

xmlFile = sys.argv[1]

# Set up persistence object
pol         = pexPolicy.Policy()
persistence = dafPersist.Persistence.getPersistence(pol)

# Where is the file on disk?  Make a storage object
loc         = dafPersist.LogicalLocation(xmlFile)
storageList = dafPersist.StorageList()
storage     = persistence.getRetrieveStorage('XmlStorage', loc)
storageList.append(storage)

# Capture any associated metadata
metadata    = dafBase.PropertySet()

# Unpersist the object; you need to say which object in storage to grab
persistable = persistence.retrieve('PSF', storageList, metadata)

# Cast to a PSF model
psf         = algorithms.PSF.swigConvert(persistable)

NOTE : without this line

import lsst.meas.algorithms as algorithms

I get the error

0: Message: No Formatter registered for Persistable name: PSF

because the formatter is only registered when the Persistable subclass is imported.