Except what if I say "screw teamwork" and just give 1241531121543512 agents the same job. Not working as a team, just each of them is to go do task <x>

Frank's hit it on the head, matrix rules fail when numbers of programs running matter. The correct abstraction to me seems to be just saying that dozens of individual agents doing the same job merely form agent voltron and count as a rating 7 agent, and all their programs are rating 7.

Tada.

Teamwork is cool and all, but I've tried it in game.

You can only get teamwork dice up to your own rating in the skill (forgot the reference) and an Agent can only have up to it's rating (the skill in this case) in a program. So an average agent, rating 3, can have a max of 9 dice in this way.

Mathematics aside, you can buy 2 successes with that, no matter how many agents you have backing the original. Probing can get you an opening eventually, but you still have to have the stealth to get in when you do. As noted above the target in that case is the Stealth rating and not a test result so no help there.

In true to game tests my hacker figured out that using a three man agent team to Exploit or perform simple hacking tests was ideal.

There was talk of setting up a virus, a rating 6 agent that went out and Exploited anything it touched, copied itself to that node then moved on from there at whatever rating it could. Trick there is I don't believe an Agent can launch an Agent. So the hacker had to manually activate each dormant Agent remotely. It was sloppy, it was slow and it wasn't worth it.

He also ran a battery of commmlinks he calls the Barracks. These were dedicated to Agents that piggybacked through his primary matrix connection and acted as his soldiers in the field. It basically amounted to what the books states. As a GM I upped the IC and had the other side call in reinforcements. To him, it felt like a WW2 sim everywhere he went.

I never had an issue with the rules. I think theory might have gotten a bit out of hand here. My players have kept to the spirit of things. I assume that if they don't mention the common problems of agent swarms in the books, they don't have them. I might be uncreative but it's never come up, and my TM reads dumpshock.

I practice it all makes sense, it works and if you don't think so I'll let you know the next time we run.

Well the aim of forcing teamwork on agents is to diminish the power of an agent horde.

Maybe you don't have problems with your players, but as stated before the real problem is with the matrix defenses of a corporation with near unlimited funds. If you have 1000 agents on a node permanently scanning for intruders and engaging those they detect, there is no way a hacker will get through that node.

Armies of agents...
- teamwork is for metahumans. I won´t even allow agents to support a hacker - any semi-intelligent tools that might be useful are part of the program used.

- agent hordes are useless on attacks, but very strong on defense. The corps should not do this because it is not necessary - most hackers are deterred by a few token ICs. That leaves the defensive setup for rigger comlinks (just an example). No problem, just a bit expensive.

- doctor smith is easy to fix. One could rule that only one healing test per subject per turn is allowed.

- Frank: If you want to fix agent hordes per unwired, a new kind of attack program targeting all agents/ICs on the node might be an option. Could be justified by overloading certain SPUs to the point of failure. (Maybe this would work as an active suppression program (agent ratings -x, failure if <=0), with the downside of automatically alerting the target node. It would also need to be run once per supressed node.).

Ryu, but there is no cost associated with upkeep of an agent horde. With security hackers, you need to keep paying them year after year, and all that. An agent hoard is good until the next matrix crash.

hmm, limited use. may be on to something there.

a spirit runs out of time and/or services, a drone runs out of ammo (and im guessing that arsenal will introduce fuel rules), but agents keeps on going until crashed. and then its just a matter of loading them again like the software that they are.

and when cracked, you can load as many as you have hardware to run them on. unlike spirits where you have to pay summoning costs, or drones where you either get a new one, or pay for repairs of the old.

so yes, SOTA rules are more or less required to keep a agent army in check.

There are a lot of ways to explain why agent armies aren't possible (if you want to go that way).
One I like is to say that lots of agents means too much checks for legitimate users, too many false positives, but above all redudancy with agents scanning one another.

If you've got 2 users but 500 agents, the agents will spend a lot of time scanning each other (which is required to prevent a hacker/hacking agent from hiding as a legitimate ICE) and the users might be able to get away unnoticed, whereas 2 users with 1 agent will have the agent scan each user for sure.
So you scale the number of agents with the number of users.

By the rules, there is no reason to not have the agent horde.

Still the corps won´t invest heavily into something they don´t need. PC hackers tend to be experienced specialists with good equipment, not the kind of opposition most matrix security will be concerned with. And upkeep causes costs, even if the game does not have rules for it. So the corp will be interested in low maintenance security.

If one or two ICs do not deter the intruder, shutting down the system is cheaper than swamping it with agents. Even complex sites can - if build the right way - severe their matrix connection by shutting down a few select nodes.

Oh, yeah, I forgot that it only works with the way I consider Matrix perception and ICE checks.

False positives can happen if an agent glitches his perception check (and with 500 agents checking each time someone's connectnig, it might happen frequently).

So what? Every couple of times a legitimate user logs in, and gets booted off, reboots his machine and tries again. Write it off as "those damn computers". Plenty of IT people currently have that attitude.