Transcription
We're back with another huge update to Klein. At this point, the Klein devs are truly pushing to make this the best AI coding agent that's available. With the amount of work they're putting in, for those who do not know what Klein is, it's an autonomous coding agent that runs right within your IDE. That's capable of creating, editing files, executing commands, browsing the web, and so much more autonomously and with your permission each step of the way.
This week we're getting a massive new update in Klein, which is Klein 3.9 and the 3.69 release. This update focuses on accessibility and performance, which is introducing a completely free client API provider option, new checkpoints, free APIs, additional providers, and much more. Let's now get started and take a look at all the new releases apart of this update.
Starting off with the client API provider. Now guys, don't get scared; it's still free, but this is a new provider option that they've introduced. And essentially, it's something that you can get started for free and get immediate access to different advanced AI providers or models, like you would with other providers like Router, Open Router, or many of the other ones that are listed within Klein. This is just the way for them to, I think, make money with Enterprise as well as with their own provider. You can still bring your own API from other models like Anthropic as well as OpenAI. They actually give you a couple of free credits to use state-of-the-art models like Claude 3.7, Sonic, completely for free without you even uh inputting any sort of credit card.
Businesses run on knowledge, but what happens when key employees leave and information is lost? Well, that brings me to introduce Sensei, which is today's video sponsor. Sensei is pioneering RepAs a Service with AI-powered digital clones that capture, extend, and share expertise. Sensei ensures institutional knowledge never disappears, powering seamless onboarding, automations, and global collaboration. Free your team from repetitive tasks and unlock productivity with Sensei's AI replicas. From knowledge retention to business automation, Sensei helps you scale, innovate, and grow. Get started today with 10% off using the code AI10 at checkout. PR Sensei now, where knowledge meets innovation.
Next up, we have Klein's latest update, which is going to optimize using the branch per task strategy. Which is significantly going to reduce storage requirements and speed up the loading time of the first task. Instead of storing redundant data, it's an approach that creates lightweight branches for each task, and it's going to improve the responsiveness as well as the efficiency of the new code that is generated. And it will easily let you track all the different changes that are made.
Another new update is that Klein introduces a toggle in settings that allows users to disable automatic model switching between plan and act modes. By default, this setting is disabled for new users, giving them more control over how the agent operates. This is going to ensure a more predictable experience while still allowing advanced users to enable dynamic switching if needed. You'll notice this when you head over to the settings tab. You can see that you have two separate modes: you have the plan and act mode that you can configure a separate API key or a model for. So, for example, if you want to use a reasoning model, you would use like the DeepSeek model for the planning, and then for the ACT model, you can use Anthropic's Claude 3.7 to focus on the coding output. And within the settings, you can obviously toggle this so that you can use different models or plan and act mode, or you can disable this so then you can just use one singular model for both.
What's also really nice is that they've added a lot of new different providers, such as the GCP Vertex AI, where you can now select the cloud region, and then you can select different models. Now, in terms of the new Quen model, you can actually go over to Alibaba Quen, and then you can search up the new Qwen model, which is something that you can access now within Klein. On top of all this, they fixed a lot of bugs, which is really nice cuz it improves the overall platform even further, and and it makes sure that there isn't any sort of inefficiency in the generation.
Now, if you used Klein before, there's a feature that you've probably seen, which is Klein's memory bank. And this is something that boosts project efficiency by maintaining contexts across tasks, which is going to reduce hallucination, rework, and overall token usage. And just recently, they have improved it even more so that it's refined to provide more efficiency. Instead of just optimizing individual prompts, it prioritizes on long-term accuracy and productivity. So think of it as a tool that you can use to save more on your token usage, which is going to reduce your cost, and it's going to exponentially save you more time in terms of getting you the most recent generation from prior context.
And I was just testing out the memory bank feature where I previously had developed the CRM dashboard with Klein, and I was able to use that context from previous generations with this prompt where I told it to find a bug, and then afterwards I simply told it to create the CRM dashboard, which used the prior context of generating this dashboard from another prompt. And it saved me a lot of tokens to develop this cuz initially I think it took around $3, but in this case it used only 99, including this previous prompt which was to fix the bugs. And this was with using the Claude 3.7 Sonic model, and it developed this CRM dashboard, a really basic one, but you get the gist of it; it was able able to do it quite easily with the context that it already had.
I should have said this before, but if you're looking to get started with Klein, you can easily install it for various IDEs like VS Code, Cursor, or Windsurf. You can go ahead and install it through their Marketplace by simply going ahead and clicking on this install button, which is linked; I'll leave in the description below. Something new that you can do by selecting the API provider OpenAI compatible is now going over to model configurations and setting different temperatures, meaning that you now have the ability to have it output based off the context or configuration that you set within the settings tab.
Another really tiny update is that if you get over to the history tab, you can actually delete all of the history you have over here. I wouldn't recommend doing this, but you can delete everything rather than going individually and deleting the history you have within Klein. Now, something cool that I've noticed is that they've added more categories of different MCP servers. Now, if you do not know what MCP servers are, I have a video that talks about it, and I actually showcase how you can develop your own within Klein, but you can see that the team is slowly adding on more from different categories. And these are just additional tools that can enhance your AI with generation, so it's something that I truly recommend that you watch cuz it's going to get you a better understanding of what you can do with MCP servers. And huge props to the team for continuously adding more.
But that's essentially it for today's video on Klein. I'll leave all the links in the description below. Huge props to the team for continuously working on this even further. I'm excited to see what 3.7 is going to be about, but with that thought guys, thank you guys so much for watching. I hope you enjoyed today's video. Make sure you subscribe to the newsletter, follow me on Patreon, follow me on Twitter, and lastly, make sure you guys subscribe to the YouTube channel. This is a great way for you to stay up to date with the latest AI news, so turn on the notification bell, like this video, and take a look at our previous videos. But with that thought guys, have an amazing day, great positivity, and I'll see you really shortly. Peace off.