Feature Requests

Using a model like Claude 3.7 Sonnet usese a lot of flow credits on tasks (like pulling in context) that can much cheaper be done using free models like Cascade Base or Deepseek V3 (once its tool usage gets better) Build in a feature to allow-list certain types of tool calls to automatically use Cascade Base instead of the selected, expensive, model. This would vastly reduce flow credit consumption by models like Claude 3.7 Sonnet

Unused Premium User Prompt Credits and Premium Flow Action Credits roll-over to the next month

deepseek v3 0324 update: Performance greatly improved, hoping it can support new models.

Ocassionally cascade will just stop performing a series of tasks. As an example, I'll ask it to write some tests for me. It'll write them, and then find some issues, fix those issues, and try again. Then it'll notice some more issues, tell me it's going to fix them, and then just fix them and stop. If I say "continue" it resumes the task. It's mostly annoying, but also breaks flow. If cascase is automatically running all of the tasks it's doing, I can task switch while it works to the result I want. But when I have to continually tell it to "keep going" I have to pay attention to it.

Your credit system is ridiculous, and it's stopping me (and probably may others) from using this. Just flat rate and give me access to all models with my own API key.

Generate Commit Messages from Committed File Context

Choose which model Codeium uses and bring your own API keys

Now, models are pulling context, using memory, using MCP, using global rules, solving lint errors... especially if you're coding typescript, you're absolutely cooked... i used 250 premium flow action credits in one day with 48 prompts. at this rate, ill have to buy 10$ credit add-ons every 7 days & it makes no sense at all! I get their business and pricing model, but it needs a serious update. we need 2500 credits on the pro plan at the very least for the same price to last a full month and get what we paid for... Assuming we are coding 8hrs a day. We also need to fix these models so they run less excessive loops in case they are not providing the correct answer. claude 3.7 can literally spend 20 flow actions doing the opposite of what i asked. this is very stress inducing and quite frustrating honnestly... This was always a thing, but with how complex the system has become now, with all the newer fetures, the models just blow through your usage... it's not nearly enough. when windsurf first came out, it was fine becasue, there were not as many tools, to use this many flow credits. But now it has made the app legit unusable, because my 15$ now lasts 15 days, its been two months in a row now. And we can either choose from the pro plan or the pro ultimate plan, which is a 50 dollars jump, and im pretty sure i could blow past 3000 flex credits without trying right now to be honnest, at least thats what using windsurf has been feeling like in the last 60 days give or take. -> I would give 1000 flex credits for 10$. 500 flow actions can easily be gone within a day or two, especially now, this add-on makes no sense at all... -> we need to either give more to the pro plan or make an entirely new tier in the middle like "wind-breaker Plan" that gives like 2500 credits and 1000 prompts for 25$ and 20$ for early birds. and 3000 flow actions for 60$ is ludacrous they should get between 4500 and 6000 for this much. its a vs code fork i believe, so its not like the app is super fast or anything and we are paying for that... we are strictly paying for advanced AI structures within a vs-like environment, so please make it right on that end and deliver on the only promise you made. -> Someone suggested to use cascade base model to pull context from files and pull from memory so the chosen model is only used for extensive work like MCP, problem solving and input/output. Maybe that would help with the whole problem. Right now windsurf is coming to a point where its unusable now and very stressful to use, because every single prompt im checking "How many flow actions has this used?". I am all for businesses generating revenue, but this pricing model needs to be addressed. This is a problem linked directly to the product value and fucntionality and its efficiency/value per price. this can quickly get out of control and make people leave in droves. which seems like its already starting. And frankly, i'm next. i don't want to be, because i have been here since the start and i genuinely liked the product at first. it was very good. If usage at the very least, isn't addressed by the team within the next month they probably won't do anything about it, so i'll leave for better options. there are at least 3 other options on par with windsurf if not better right now. cursor, vs code with cline, copilot etc, its a competitive market so i won't have trouble finding a new tool/extension. The waves updates are great, but if you're not addressing the elephant in the room, people will just leave, as right now it seems like, if you are just a regular developper with the usual tech stack and workload, you are paying a MONTHLY subscription for about 10-15 days of LOCKED usage. this is so bad this can really kill the product if most people can't use it past 15 days you know...

Claude 3.7 burns through flow credits like nothing else. I'd say 5-9 flow actions per prompt for me. That means that 3.7 can't be used all that much before credits expire severely limiting the usability of 3.7 in windsurf. That needs fixing. The model is good from what I can tell but I went from low flow credits to one third used in a single session.

When will xAI Grok 3 be available as a model?