Transcription
[Music] [Applause] Hi, welcome to another video.
So, Rue Code has been on a roll with a bunch of updates lately, and today I'm going to walk you through everything new from version 3.20 all the way up to 3.25. If you're using R code or thinking about it, this will give you a clear sense of what's changed and what you can actually do now that you couldn't before.
Let's start with R code 3.20. This was the release where the experimental marketplace for extensions and modes first appeared. That means you could browse and install community created model context protocol servers and custom modes right from rue code with no experimental flags needed. They also introduced concurrent file edits and reads, which is pretty useful if you're working with lots of files or big code bases. Another thing was prompt history navigation. You could use the up and down arrows to cycle through your prompt history when your cursor was at the start of the input. That really streamlined the workflow.
Moving on to 3.21. This was when the RU marketplace officially launched for everyone. You could now discover and install MCPs and modes with just a few clicks. They also rolled out direct support for Claude Max subscriptions through the Claude code provider. So you could use Sonnet 4, Opus 4, and other models included in your plan without API keys. Excel file support became more robust, and you could render latex equations in the chat.
Now 3.22. This was more of a patch release, but still had some significant changes. Timer-based auto approval for follow-up questions was introduced, so R code could proceed automatically after a timeout, reducing manual intervention.
Jumping to 3.23. This was the release where codebase indexing graduated to a stable feature and became always on, always ready. Once set up, the indexer would run in the background, keeping your semantic project understanding up to date. A new task to-do list feature arrived, displaying a checklist of steps for your task directly in the chat, which is great for tracking progress. This is similar to what you see in Claude Code as well. Command deny list was added, so you could automatically reject unwanted commands in workflows. Root code cloud weight list signup became available both on the website and in the extension, which we'll go ahead and talk about more going forward. Claude code provider got native Windows support, and command validation logic got smarter, only autodenying subshell commands with denied prefixes.
Onto 3.24. This introduced hugging face as a provider, so you could access thousands of open-source models directly. There was also the agent rule standard with agents.md files, which made it possible to share natural language guidelines across ruode aderle and other compatible tools. Terminal command permissions got a new interactive UI, letting you manage command patterns visually with intelligent pattern suggestions and quick toggles.
Finally, 3.25. This was a big one. Custom slash commands were introduced, letting you automate workflows by placing markdown files in the RU commands folder. There's a management UI for creating, editing, and deleting commands. And you get argument hints and rich descriptions for self-documenting commands. Message queuing was added, so you can type and send messages while RU is still processing earlier ones. On the provider side, Cerebras, Dubao, Sambanova, and new shoots AI models were all added. Also, there's a new thing called Ruode Cloud as well, which basically extends your R code extension with cloud-powered features for enhanced collaboration, data persistence, and usage tracking. It allows you to view your task history online, share collaborative features, and monitor your AI assistant usage and optimize costs.
Now, let me show you how all of this works as well. But first, let's talk about today's sponsor, Dart. Tired of juggling tasks across different tools? Dart combines traditional project management with powerful AI features that actually get work done. Beyond organizing tasks and boards, Dart's AI can brainstorm project ideas, generate task lists, and even complete entire assignments for you. Their composer-like AI agent understands your full project context, so you can simply chat with it to create, edit, or delete tasks naturally. The real game changer is the custom agents. You can create custom agents that trigger from the built-in integrations or a N8N workflow or custom web hook for full customization. You can create a coding agent that pushes pull requests to GitHub, a marketing agent for campaigns, or a mailing agent for outreach. Then just assign tasks and watch them get completed automatically. Plus, Dart integrates seamlessly with your existing workflow through their MCP server, connecting directly to Claude, Chat, GPT, and other AI tools you're already using. Most features are completely free with premium options starting at just $8 per month. Check out Dart through the link in the description. It might just transform how you work.
Now, back to the video. Let's start in the marketplace. If you click the marketplace icon in the top bar, you'll see a catalog of available MCPS and modes. You can filter by category or search directly for what you need. For example, if you want to add Brave Search as an MCP, just click install, choose whether it's for your current project or globally, and enter any configuration details like API keys. Once you hit install, the MCP is ready to go. Installing modes works exactly the same way. Pick a mode, adjust any settings, and you're set. Everything you install shows up in the relevant tabs, so you can easily verify what's active. The coolest thing here is how seamless it all is. You don't have to mess with any config files or restart the extension. Just point, click, and you're done, which is kind of cool.
Now, let's talk about codebase indexing. With the new always-on indexer, R code analyzes your whole project and generates semantic embeddings for all your files. This means the AI can understand not just individual files, but the structure and relationships across your codebase. When you ask Rue to find something or make a change, it queries the vector index to get the most relevant files and context, then uses that to drive its actions. This makes searches faster and more accurate and really boosts the quality of code suggestions and refactors. You'll notice that even for large repos, Rue Code stays responsive and doesn't get lost, which is quite awesome.
Next up, the to-do list feature. Whenever you start a new task, Rue Code can break it down into a checklist of steps, which shows up right in your chat window. As the AI works through each step, it checks them off, so you always know what's done and what's left. This is super useful for keeping longer tasks on track and making sure nothing falls through the cracks. You can enable or disable to-do lists per model in the advanced settings. So, if you've got a model that's not great at multi-step tasks, you can turn it off just for that one. This kind of flexibility is pretty good.
Let's look at custom slash commands. This is one of my favorite new features. You can create your own commands using simple markdown files. Just open the command manager, pick a name, and R code generates a starter file for you. Fill in the description, write your prompt, and add any argument hints you want. When you use the slash command in chat, R code shows you your argument hints as visual placeholders, making it super easy to remember what goes where. These commands are reusable, easy to share, and can turn any multi-step workflow into a single action. For teams, it's a huge productivity boost and helps standardize best practices. And if you ever need to tweak a command, just edit the markdown file and you're done. It is very similar to how Claude Codes workflows work.
Now, about concurrent file edits and reads. If you're working on a big project, Rue Code can now edit or read multiple files at once. This saves a ton of time if you're doing sweeping changes or need to gather info from lots of places. The system is smart enough to handle all the underlying complexity, and you can prompt it to do multifile operations directly. It works especially well with models like Opus 4 and Gemini 2.5, but you can tune it for other models as needed.
Another thing I want to highlight is the new support for images in the read file tool. You can now point R code at a folder full of screenshots or scan documents and it'll extract text using OCR. This is awesome if you're dealing with documentation, receipts, or any kind of visual data you want to pull into your project.
Now, if we talk about the R code cloud features, then you'd have to first sign into ROC cloud by hitting this icon. And this will take you to this site. This is RodeCloud. And now you can see the usage dashboard of the stuff that you did, as well as you can see the history of the tasks that you have done with it. You can see the model that you use and everything. You can also share the task with anyone else if you need to do that. You can also get a public share link from the R code extension itself, which is also awesome.
Another thing that I want to talk about is the free APIs. So currently I've been using the Horizon beta model, and it is pretty awesome and it's free, fast, and without any limits. So that is something that I have been using. Another thing that I have been using is Quen 3 coder free API via open router and Cerebras. They both have free API for it with different limits. Secondly, Cerebras has launched a plan called Cerebras Code with $50 and $200 plans that give you access to Quen three coder API with almost unlimited access, similar to Claude Code, while giving you 2,000 tokens per second speed, which is quite awesome. And you can give that a try, or if you have tried it, then please leave a comment if you like it or not, as it will help others. If you want more models for free, then you can also check out Kilo Code, which has about $20 of free credits and allows you to use any model. It is a fork of Root Code, and you'll find all the features you find in Root Code there as well, which is also awesome.
That is majorly all the new updates. You can go ahead and give this a try because all these new features are pretty awesome. You should definitely use this. And I use it as my main coder over things like Claude code or stuff like that. Overall, it's pretty cool. Anyway, share your thoughts below and subscribe to the channel. You can also donate via superthanks option or join the channel as well and get some perks. I'll see you in the next video. Bye.
[Music]