They still need a reason, meaning that it cannot be purely self-caused, and that reason is always traceable to external factors. At the very least, this theory represents a causal oversimplification.
This is the part of agency theory that's really unsatisfying to me, but also makes me feel like I just don't understand the theory properly. But I'm wondering if the following might be a little more satisfying to you.
So you're right that an agent needs a reason to act and that this reason can be traced to external factors. The agency theorist, I don't think, would deny this. The difference is that the external stuff is an explanation of the agent's mental state. In other words, the relationship here is an explanatory one, not a causal one.
What's the difference? Well, suppose that Sarah has a crippling fear of water. She won't go near a pool and you can forget about taking her to the beach or on a boat. It ain't happening. Sarah's fear comes from something that happened to her when she just 3 or 4 years old. She was riding in her father's boat when she fell overboard. Fortunately, she had on a life jacket, but it scared the absolute crap out of little Sarah. To this day, she feels anxiety when she even looks at a picture of water as those extreme emotions start to come back.
What we have here is an explanation of Sarah's mental state (or, at least, part of her mental state). We can also explain why she would decline to go on a fishing trip, for example. That terrifying incident explains why she doesn't want to go fishing. But that incident didn't cause her to decline to go fishing. Instead, the stuff inside her skin and skull caused that decision. And that's precisely what agency is!
Now, you might want to push the point here and say that the boat incident did cause her to not go fishing. We look at her mental state, her desires, and track what caused those desires. Ultimately, we trace those causes back to the incident in question. So, it's still part of the causal chain.
But suppose that incident never actually happened. Suppose that she dreamed it when she was little, but the dream was so terrifying and so visceral that it impacted her significantly. At such a young age, she was unable to distinguish the fear that came from a dream from real-life danger and this dream somehow made its way into her psychological makeup. (This, by the way, isn't a far-fetched story. In fact, I've spoken with people who have had a fear of something precisely because of this kind of scenario. What's interesting is that they can't shake that fear, even though they have since learned that the incident never actually happened.)
With this modification, it looks like our causal story breaks down. An event that never actually happened can't be the cause of some other event--that would be silly. Now, you might say that Sarah's dream is the cause of her fear. Story over. But this dream is something that took place inside her skull. It's now an internal cause. And what's more, it's one that doesn't seem to have an obvious cause.
There are still objections to this story as I've told it. But let's step back for a second and look at what agency theory says and, for that matter, what agency is. To be an agent entails moral responsibility. It also entails intentionality, desires, and certain mental states. So the agency theorist is looking at agents at being self-efficacious. That is, agents can make themselves do things (even if they don't really want to). The upshot is that it looks similar to the Humean move of stopping the causal chain at the point of desire. But now, it seems like we have a principled response as to why we stop there. Since agency is the sort of thing that is self-efficacious, there's no need to look beyond the internal stuff as part of the causal chain. In fact, everything outside of you (for the most part, at least) bears an explanatory, rather than causal, relationship with an agent's actions.