yes please

OMG Yes! You would take all the market share from Void basically. I get the point that your custom model SWE1 is really good too -- the best one for coding in my opinion. But Void's promise that using your own token "gives you control of sending your data to your own private LLMs" will disappear if you implement this.

Wake me up when BYO Agent is working. Meanwhile I'll be using VS Code with my own Claude 4 Sonnet

I am going broke w windsurf 😞

Projects like Void Editor are going to cause issues if this isn't implemented soon. Void is getting really close to comparing to Windsurf and its open-source and free and only in beta with a wide-range of API's already added with local vLLM and Ollama support. Not to mention they have Ycombinator backing them.

I would happily pay my 70$ a month for the base use and delegate compute to cheaper models. I dont want to sustain 200-400$ months on a single software. Please offer bulk pricing or individual API based model selection.

One day RooCode or Cline will be as good as Cascade, then what?

Not all work need Sonnet, some can be done with cheaper/free models, that can be ran on local machine. We are developers FFS, we have good machines. Will pay subscription, will use your Sonnet but my local LLM can do most of job. Like Junior/Senior pair!

The issue is that it will probably be like Cursor and all the useful features are disabled if you only use your keys. They want you to buy credits when you run out. I find that using my own keys is cheaper for me in the long run but I don't want to make my own IDE so I pay the 10 bucks a month.

I would also like to be able to use my own API keys

Your Cascade "Agent" is great and from what I've seen thus far, I'd gladly keep paying a subscription for it to continue driving and managing memory of codebase and context for whichever LLM via cloud or local Ollama API calls I choose to connect to.

I'm using OpenHands and Cursor to achieve this currently but the Windsurf interface and using the Cascade Agent as a prompt interface is far superior, so make a custom API setting, lighten the load on your servers and give us the feature we need.

When we use our own API keys in Cursor, we lose the ability to access the Compose prompting without a subscription, I have not been as impressed with Cursors underlying agent as with Cascade which is why I would be happy to keep paying your subscription however this may be the technical challenge you are facing with using external API keys since the Cascade agent most likely has a specialized way to interact with the models you offer for credits.

I would like to setup several custom API key settings and be able to switch between them as needed, sometimes I just want haiku 3.5 to do some project initialisation quickly and then I move to larger models for more complex tasks. We create several API keys for each cloud service and model, we are able to do better cost comparatives when reviewing logs and tabulate if a cheaper model actually cost more due to errors and bad loops which we cannot accomplish using your current credit system. We need more insight into how many repeat calls were made and how big the context is, custom API settings allow this.