I guess I should elaborate.
The only way that the agentless HTTP metric works is if the URL you want to monitor is:
http://YourMachineNameAsItAppearsInTheInventory/WhateverPathYouSpecifyInTheMetric
Which is, of course, next to useless.
If you're in a situation like 99% of the world where you have multiple sites hosted on a server, or if the URL for the site is not the name of the machine it's hosted on, then you're out of luck.
One of the biggest frustrations with the product is that there's no real connection between inventory and monitoring. You may have an inventory of all of your IIS and Apache virtual hosts, but there's no easy way to tie that into monitoring.
The two different methods I mentioned earlier work like this.
Let's say you have a server named ATLWEB01, that hosts a site named www.mycompanywebsite.com.
You could create command metric named 'www.mycompanywebsite.com availability'
with a command line that looks like this:
wget --tries=1 --timeout=25 -q -nv --no-check-certificate --header="Host: www.mycompanywebsite.com" -O www.mycompanywebsite.com.OUTPUT https://www.mycompanywebsite.com/Welcome.jsf"&&grep -i -m 1 -c SOMEPHRASE www.mycompanywebsite.com.OUTPUT&&rm www.mycompanywebsite.com.OUTPUT
This will use wget to hit your site, grep the output of the page you're looking for for SOMEPHRASE, only check for the first instance of SOMEPHRASE, and then return the count of SOMEPHRASE that it found. So, if the site is available, and the phrase exists, you'll get a '1' and if it's down or the phrase isn't on the page, you'll get a '0'
Then we create a rule named 'www.mycompanywebsite.com is DOWN'
and we have it evaluate our metric and see if the value is equal to 0.
You'd then create a policy with that rule in it, and as a target for the policy, point it to a server that has wget installed on it (pretty much any linux server, or any windows machine with grep and wget for windows installed (http://gnuwin32.sourceforge.net/packages/wget.htm))
Or..
You could use another monitoring tool that can send SNMP traps to the NS, and just have it all sent to the event console.
You can see where the command metric route it tedious if you have LOTS of websites to monitor.
Best of luck!