I was thinking about this topic because it's a foundation issue for good GMing, and for good roleplaying in general. Agency is one of those bits that gets tangled up with a lot of other things. It's small when you give it a passing glance, but the more you look at it, the more important it becomes. If you're really rocking a game, you can weaponize agency into a powerful tool - and indeed, a lot of modern games do exactly this.
Agency is, put simply, the power that is shared amongst players at the table to create the ongoing narrative of the game. In the traditional gaming world, much of this lives at the level of the player character and how it interacts with the rules. The more accomplished the character becomes, the more mechanical advantages it has to affect the world of the game, mostly in the realm of combat, but often in less obvious or flashy ways. Agency also lives at the level of social contract in a game - unless the rules explicitly define what power players have (assuming a model where there's a GM at all), the GM decides where the players can make contributions. This may include players inventing world details as part of character building, such as important NPCs, or including backstory that is explicitly designed to be part of play.
An abusive GM limits or circumvents player agency completely, outside what is circumscribed by the rules themselves. Player contributions are often not included in the game as it unfolds, and in the worst cases, even the ability to choose can be taken away from the players; this includes abusive behaviour like railroading and illusionism. This kind of behaviour seems to stem from either a lack of trust at the table or the GM's lack of confidence in their own abilities. Either way, it's no fun to play a game like that.
For most gamers, being a player is all about being able to choose. If the point of the game is to explore a world, they want to choose where they go and what they see. If the point of the game is to tell a narrative, they want to be able to push the story in directions they like. If the point is dramatic interactions, they are in control of what their characters decide and what their emotional reactions are to things. Nothing else is acceptable.
The good news is that a lot of modern games, even "traditional" ones that are more interested in fighting and exploring, have begun to include a lot of rules that give players rules support for greater agency. In storygames, often player agency is very strong, sometimes equal or even greater than the GM's power. Fate, for example, presents a robust set of tools that encourage collaboration from the beginning of a game and throughout play. Players can easily introduce new angles into a story that the GM can't see coming, as easily as creating a new Aspect.
This is good news because collaboration makes games better. I like games with a strong GM role, but I expect my players to take the reins and make interesting choices. I hold my story ideas lightly, or use them as a colourful background for the actual, important plot events that flow from player character choices.
In a dramatic game, player character choices are the most important part. Everything you are doing is about presenting the characters with important choices, meaty choices that can have significant implications for the outcome of the game. These kind of choices have interesting fallout no matter what way the player decides to go. (For those keeping track, yes, we are talking about Kickers and Bangs here.)
If the players have agency at the level of calling for scenes, you're really getting into tasty territory. Now the players can jump right to the most urgent issue for their character, dealing with it much earlier and more directly than a traditional narrative structure would allow. Decision, fallout, the story moves in a new direction. All independently of what the GM would have planned.
This is great because emergent play is exciting for everyone involved. A lot of traditional gamers feel that games that are aiming at story or character-focused play are somehow railroady or less pure because there is a level of explicit preparation and structure that goes into a good game, but it's actually more like setting up bowling pins together and then knocking them down in interesting ways. There is still a high level of emergent play in those games, it's just pointed in a specific direction, as opposed to traditional games that may have those same things emerge over the course of long-term play. Sure, it may be a powerful moment in a five-year campaign when your character renounces his family name and walks away from House Gygax, but in a dramatic game you'd start a lot closer to that moment. The game would be about that moment. And even if your character made the same decision in session one, there would be plenty of interesting play to come.
Players in dramatic games are usually explicitly collaborators in the process of making dramatic moments happen, and they often have a high level of agency in creating scenes and moving the story along. GMs used to a strong hand at the table may find it a little scary to have so little control over the game, but I find it liberating. When I sit down at the table, I've done my preparation, but I know I'm not carrying the whole burden of making the game interesting on my shoulders. Agency makes the players my partners in that effort.
Agency makes me, as a GM, feel more like a player in my own game and less like a button the players push for story or adversity.