Configurable Key Bindings (Enabling Fast Model Switching)

[youi-architect]

Describe the feature or problem you'd like to solve

I'd love to be able to bind keys like F1, F2, F3, etc. to slash commands. The main reason is faster, easier model switching. At the moment, switching models means using /model and typing memorized model IDs, which adds just enough friction that I do it less often than I should. The same pattern could also be useful for other slash commands such as /new, /resume, user-invokable skills, and similar frequent actions.

Proposed solution

Let users map function key bindings to slash commands via something like a ~/.copilot/keybindings.json file.

For example:

{
   "keybindings": {
     "F1": "/model claude-haiku-4.5",
     "F2": "/model claude-opus-4.6 xhigh",
     "F3": "/model gpt-5.4 medium",
     "F4": "/new",
     "F5": "/resume"
   }
 }

The main thing for me is that this should work while I'm already typing a prompt, without losing the draft input, as that is usually when I decide which model would be best for it. That means a single quick keypress after typing my prompt could be equivalent to:

1. <ctrl+s>
2. model
3. <enter>
4. ↓
5. ↓
6. ↓ (to the appropriate model)
7. →
8. → (to the appropriate reasoning effort)
9. <enter>

I also appreciate this would be less clunky if I could control the reasoning effort inline through /model - for example something like /model gpt-5.4 high (that is, /model [MODEL] [EFFORT]). That alone would be a great improvement, but when the full model choice can be expressed in one command, it can also be mapped easily to a key binding.

Example prompts or workflows

The real value would be in mapping slash commands that get used frequently:

• I'm in the middle of a session, want a quick, cheap answer, so I press F1 to quickly switch to my predefined lightweight model
• I type out my next prompt and, after reading it back to myself, I realize it should have some really deep reasoning behind it, so I press F2 to switch instantly to my predefined heavy-lifting model with xhigh reasoning effort
• F3 runs /review
• F4 runs /new
• F5 runs /resume

For a more polished finish, it could also be nice to have a /keybindings command for maintaining these mappings, similar to /mcp.

Additional context

I should probably switch models more often than I do, but right now there is just enough friction that I usually don't bother, meaning I end up using higher compute models for basic tasks and I end up staying on the same model longer than I should, even when a cheaper or faster one would be better for the next turn. If switching models is fast enough to become muscle memory, it is far more likely to be utilized.

Whilst full 'Auto' mode would would also be useful to help manage premium requests more efficiently, this functionality gives power users full control to quickly choose the model configuration that suits them best.