There's an interesting discussion of 'agent causation' accounts of free will at "OrangePhilosophy" (yet another philosophy weblog) by Jeremy (the last name isn't given at that point, but I'm fairly sure it's Pierce). The problem proposed is an interesting one. The most reasonable move by the 'agent causation' people, I think, would be to deny that events, properly speaking, are either causes or effects. In other words, causation is not a relation between events; the only relations between events are relations like spatial and temporal contiguity (before, after, overlapping, etc.).
Nor would this be an unreasonable move. This whole 'event causation is the only causation' fad was, one could argue, started (unintentionally) by Malebranche in the 17th century. He did it in order to attack what he considered to be the idolatrous notion that creaturely substances could be true causes. To put it in other words: event-based accounts of causation come into their own when people starting denying that the earlier (largely scholastic) accounts of causation in the created world. Malebranche still held that this older account of causation was the account of true or real causes; creatures were not true causes but occasional causes, i.e., 'causation' for creatures consists of events organized by laws governing God's activity. He gives a number of arguments to supplement his conclusion, the most famous being the case of billiard balls hitting each other on a table. (Malebranche, we are told, was very good at billiards.)
Berkeley takes up something similar, but extends true causality to all spirits (he is very explicit about this move). I suspect it is from Berkeley that Reid originally picks up his notion of an agent, although I don't know for sure. (This would be a good thing to research; I'll have to think about it more.)
Hume does away with what Malebranche and Berkeley considered true causation, and just keeps what they considered to be the derivative causation of creatures (Malebranche) or sensible objects (Berkeley). (Hume, of course, is closer to Berkeley's version, but he uses a number of Malebranche's arguments, including the billiard ball argument.) He also, I suspect (there needs to be more research on this), is the one responsible for the rise of popularity in discussions of causation in terms of events (although he wouldn't have been the only one). Analytic philosophers tend to accept the event causation = causation view because they are uncritically Humean on this point. Unlike Hume, however, they don't realize how limited the set of kinds of relations events can have to each other really is. Hume's recognition of this is the key to the entire Humean account of causation. What makes the Humean account of causation so strong is that, if you once start thinking of causation entirely in terms of events, you are pretty much committed to the path Hume takes, unless you inconsistently splice your 'event causation' with elements of (e.g.) an 'agent causation' account.
People who advocate what are commonly called 'agent causation' accounts of free will usually don't take this road; they more or less accept the event causation orthodoxy but add another kind of causation on top of it. This need not be inconsistent; it can only be inconsistent (to return to the author's argument) if event causation is fundamentally deterministic. This is another issue with which I have problems for related reasons, but I'll leave those out. [[LATER NOTE: I just realized that the first sentence is vague; "this road" is the alternative of denying that 'causation' actually applies to events.]]
Jeremy has a closely related post at "Parableman," also worth reading. Actually, I find I like a lot of Jeremy's posts, so I'll probably put his weblogs up on my weblog list, as soon as I get around to it.