diff --git a/utils/tko_publish.py b/utils/tko_publish.py
new file mode 100644
index 0000000..269dd02
--- /dev/null
+++ b/utils/tko_publish.py
@@ -0,0 +1,110 @@
+#!/usr/bin/python
+"""
+This script will scan an autotest server results directory for job result
+directories that have completed and that have not yet been published on
+a remote dashboard server matching given filtering options and for those it
+finds it will rsync them to the tko server and mark them as published (it uses
+a <jobdir>/.tko_published flag file to determine if a jobdir results directory
+has been published yet).
+"""
+
+import sys, os, re, optparse
+
+import common
+from autotest_lib.client.common_lib import utils
+from autotest_lib.server import frontend
+
+options = optparse.Values()
+
+USAGE="""tko-publish [options] <resultsdir> <rsync-destination-path>
+
+Where:
+<resultsdir>              A path to the directory having the job results
+                          directories to publish.
+
+<rsync-destination-path>  A valid rsync destination path where to upload the
+                          job result directories.
+                          Example: user@machine.org:/home/autotest/results"""
+PUBLISH_FLAGFILE = '.tko_published'
+RSYNC_COMMAND = 'rsync -aqz "%s" "%s"'
+
+
+def get_job_dirs(path):
+    regex = re.compile('[1-9][0-9]*-')
+    jobdirs = []
+
+    for dir in os.listdir(path):
+        # skip directories not matching the job result dir pattern
+        if not regex.match(dir):
+            continue
+
+        dir = os.path.join(options.resultsdir, dir)
+        if (os.path.isdir(dir)
+                and not os.path.exists(os.path.join(dir, PUBLISH_FLAGFILE))):
+            jobdirs.append(dir)
+
+    return jobdirs
+
+
+def publish_job(jobdir):
+    cmd = RSYNC_COMMAND % (jobdir, options.dest)
+    utils.system(cmd)
+
+    # mark the jobdir as published
+    fd = open(os.path.join(jobdir, PUBLISH_FLAGFILE), 'w')
+    fd.close()
+    print 'Published', jobdir
+
+
+def main():
+    jobdirs = get_job_dirs(options.resultsdir)
+
+    afe = frontend.AFE()
+    # the way AFE API is right now is to give a whole list of jobs and can't
+    # get specific jobs so minimize the queries caching the result
+    finished_jobs = afe.get_jobs(finished=True)
+
+    if options.jobname_pattern:
+      jobname_pattern = re.compile(options.jobname_pattern)
+    else:
+      jobname_pattern = None
+
+    # for each unpublished possible jobdir find it in the database and see
+    # if it is completed
+    for jobdir in jobdirs:
+        job_id = int(os.path.basename(jobdir).split('-')[0])
+        job = [job for job in finished_jobs if job.id == job_id]
+
+        if len(job) != 1:
+            continue
+
+        if jobname_pattern:
+            # does it match the jobname pattern?
+            if not jobname_pattern.match(job[0].name):
+                continue
+
+        # does it match the wanted job owner
+        if options.job_owner and options.job_owner != job[0].owner:
+            continue
+
+        publish_job(jobdir)
+
+
+if __name__ == '__main__':
+    parser = optparse.OptionParser(usage=USAGE)
+    parser.add_option('--jobname-pattern', dest='jobname_pattern',
+                      help='Regexp pattern to match against job names, by '
+                      "default there won't be any matching done",
+                      default=None)
+    parser.add_option('--job-owner', dest='job_owner', default=None,
+                      help='Job owner username to match against for the '
+                      'published jobs, by default no matching is done.')
+    options, args = parser.parse_args()
+
+    if len(args) < 2:
+        print USAGE
+        sys.exit(-1)
+
+    options.resultsdir = args[0]
+    options.dest = args[1]
+    main()