❓ MobX-State-Tree: Is there any elegant solution to the self limitation in actions?

[ryskin]

Hey everyone,

I’ve been hitting a frustrating wall with MobX-State-Tree + TypeScript, and wondering if there’s a clean solution in 2025 — or if this is still just “how it is.”

Whenever I define multiple actions inside .actions(self => ({ ... })), I can’t call one action from another without running into TypeScript issues. The problem is that self doesn’t recognize other actions yet during the model construction phase.

.actions(self => ({ increment() { self.count += 1; self.log(); // ❌ TS error: Property 'log' does not exist }, log() { console.log("Count is", self.count); } }))

I know the usual workarounds:
• Split actions into multiple .actions() blocks (feels messy)
• Use as any (yuck)
• Cast self manually (boilerplate hell)
• Or just stop using MST altogether and go back to plain MobX (which I’m seriously considering)

But my pain got worse when I tried using Zustand-style cursors or functions that interact with many actions dynamically — then I end up juggling tons of small action keys and casting types all over the place. It kills DX and makes the code harder to read and refactor.

❗ So the question:

Has anyone found a more elegant way to compose MST actions without fighting TypeScript and the self reference?
Or has the ecosystem basically moved on from MST when you need flexibility?

Would love to hear how others are handling this in real-world projects.

Thanks!

[thegedge]

I think there three options right now:

1. I believe this works if the actions are in the same actions block. self works for calling actions in previous actions blocks. Neither work for actions declared in future actions blocks.
2. Use a library built on top of mobx-state-tree, like mobx-quick-tree or classy-mst. Disclaimer: I built the initial implementation of that library and work for gadget 🙂
3. Find someone to complete the selfless model work.