Kilo Code: AI Coding Agent, Copilot, and Autocomplete

Super!

Best open source IDE for AI, and full featured. Works very well with their suscriptions or with others. It's real open source... I feel free!! Thanks a lot to the team.

I've tried ClaudeCode, Cline, Continue, for me this one is the best "side panel" AI agent in VSCode, it natively supports BYOK, and it is super fast that significantly boost my workflow

Model is stuck in loop and burning money very easy.It is even simple job.I just request simple code changes and It is burned a few dollars. It is impossible to reach that model spend like money. I like Kilo code concept but I was used cheapest models and It is burning money a lot.

i love it

Barely usable. I genuinely don't understand who likes this. The UI is terrible and cluttered with models I never asked for. If I connect my own provider, I want to see only that provider’s models - not a bunch of unrelated options polluting the interface. UPD: Even worse, after uninstalling the extension, the Kilo binary kept running in the background, making my MacBook run extremely hot. That’s absolutely not acceptable.

Inflexible with certain features needed by some users.

Kilo like agent is a useful but not extraordinary tool, some of the FREE MODELS are crazy BAD, beware with LAGUNA XS.2! git push unauthorized and a lot of hallucinations , deleting an corrupting parts of code. ** beware with LAGUNA XS2 dear DEV! , and I you go paying the best is CLINE

Highly useful AI agents for coding. The plugin's UI looks very nice, in my opinion. Kilo allows you to automate your coding tasks even on the free tier with free models. This plugin is awesome — I use it in Cursor and VS Code. The developers keep updating it, and it just gets better and better, even though it's already a very good and useful tool.

it's a pleasure to use it

These guys are doing really great stuff. Using frontier models will not always be the way. I highly recommend getting into this tool for working with many different models. If you can get good at programming with open weight models then you will be extremely valuable in the future. Definitely a tool worth adding to your belt.

A complete piece of crap, slow, stupid, visual glitches,

Sou iniciante no mundo code ... mas... pelo visto... o Kilo Code superou minhas expectativas... o auto completar não funciona no vscodium... mas... consegue cumprir quase todas as tarefas que é proposto.

The best pay-as-you-go agent so far and has lots of free models

Best free agents, no struglling in the initial configuration. Good paid deals in AI Agents and other features.

Works well and the free tiers are fast. Kilo team support is better than most.

Very good, and helpfull. My most used feature is Generating Commit Code

Excellent and nearly always improving. Accurate, fast, open and hugely flexible with a great community. Well done.

Nice tool, nice you

Thats Very good and hi speed in free

Very useful. And there are usually 2-3 free models at a time to allow completely free usage.

I run Claude Code as my daily driver, but Kilo Code earns a permanent spot in my workflow for a simple reason: it's the only open-source VSCode extension that treats the entire codebase as a canvas rather than a single file. The multi-file refactoring, custom modes, and 500+ model support mean I can swap in whatever model fits the task — budget-friendly for boilerplate, frontier for complex reasoning. The agent config is transparent and hackable, which matters when you want to actually understand what's running your code. What impresses me most is the team's cadence. Bugs get addressed, releases keep shipping, and the community feedback loop is real. That's rare. If you're serious about AI-assisted development and don't want to be locked into a single provider's walled garden — this is the one.

Unpopular opinion: Stop looking for standalone AI apps. Kilo is the most robust, platform-agnostic coding agent on the market. CLI navigation, deep multi-file IDE refactors, automated repository hygiene—it treats my entire codebase like a single canvas. Local-first flexibility paired with elite model weights makes it the absolute backbone of how I ship software. It’s not an assistant; it’s a co-founder." Tyrone Smith Founder, Vauntico · Profile ID: ZA-GEN-001 🛡️ Verify Posture: vauntico.com/verify/ZA-GEN-001

The autocomplete feature is a paid feature, which is disgusting.

I’m using Kilo Code for my Blender addon development. It really helps to construct my Python modular script, writing code, inline autocomplete suggestions. Kilo understand what I want to do. You can build a code even with budget friendly models and modes, no need to sell your car or house to use T-1000 model terminator brains like Claude. Every new update may include some problems or issues. However, it’s good to know there is an active development team behind Kilo whose listen the users and make Kilo beter and beter. I really enjoy my code sessions with Kilo 🤘🦋🌸

One of the last few (if not only) FREE and OPEN SOURCE VS Code coding assistant extensions that is super customizable (modes/agents and skills and much more), has a huge support community, and is constantly shifting and growing!

Cool staff and concept

Its nice, but please update and improve the UI like WindSurf!

很好用的一款ai工具

Version 7 was released to prematurely. But it does get better and you can feel improvments with every release. I still have hope that the early bugs will decrease with each subsequent release. For me still the best VSCode extension out there for agentic coding.

You had to screw everything up like that. Bad, wery bad.

v7 broke the product, too bad because it was the best coding agent available

After trying out several coding agents, I still find Kilo Code the best to use—though I'm currently on an older version.

For those who are dissatisfied: people are actively working on this open-source product, and it’s already helping a lot of users. They continue to improve it, and version 7 includes many features that weren’t present in version 5. In that context, saying that they will lose users seems a bit premature. A more constructive approach would be to clearly describe any bugs or issues you encounter so the developers can continue improving the product. And if the tool doesn’t meet someone’s needs, there are plenty of paid alternatives available — for example, Cursor — though they may not necessarily offer the same range of functionality as KiloCode.

It was the best product available. But you decided to change it... Why did you do that? You have broken your product, making it not only useless, but also DANGEROUS. He broke the whole project, all the documentation, and generally ruined everything that was there. I am horrified at the thought of how many people have been injured. Congratulations on the release of the consistently crappy version 7. You will lose all users in the coming days if you don't switch back to version 5. (sorry for the google translator)

In the latest release Kilo Code keeps getting stuck in "Making edits" loops due to the faulty file edit tool and it has been unusable. They managed to mess up a perfectly fine extension and one wonders why, really. I had invested into using Kilo Code for most of my projects and now I need to quickly migrate to another extension. For the time being I think Roo Code is the best alternative.

The newest "stable" v7 update is an absolute mess and completely changed the whole tool. If you are researching extensions: Discard any older reviews. Next to being a completely different (and aweful workflow), it introduced so many bugs and issues. Checkpointing is broken, sometimes the model selection bugs out (cost me 8$ since it accidentally selected an expensive model for a long task). Absolutely horrifying. While v5 is still usable, it doesn't make sense to support a tool that will develop into a completely wrong direction.

It’s surprisingly difficult to make a product worse at a time when AI-assisted coding has become standard. Yet Kilo Code somehow managed it. The latest update (from version 5 to version 7) has been a major step backward and has seriously hurt the overall experience. I have started looking into other alternatives.

I'm an amateur so I'm not able to give a very high-level analysis, but thus far I think the interface is good, Kilo's token pricing seems reasonable, the free model offerings have been great and it's easy to use a local LLM. I dig it.

Thank god i found out about this. I am non developer and almost my entire project is created by Kilo Code. Product is so good very soon I might just be living on subscription fees received from that product.

Thanks guys for still letting me use version 5. That’s why it’s 5 stars! As soon as version 7 becomes mandatory, it’ll be 7 % 5 stars.

This v7 is a huge feature regression that never should have left early access in the current state... TLDR: People, just stick to version 5.x or just consider roo or cline or something else until this is done cooking... I get your wants and long-term needs of unifying code between cli and vscode and such but this is no proper way of dealing with major version "upgrades" at all! - no more orchestrator, no sub-agents(?). (possibly both context bloat and context rot when using v7 over v5 but I don't have enough data to actually back that statement up other than "gut feel" for now) - no possibility to swap model or force-stop agent once a rate-limit have been hit. (it will just keep failing forever with an increasing timeout between the failures) - complete lack of many missing config options compared to earlier - no insight AT ALL at what agents are doing sometimes, no request logging, no information - reasoning blocks can mostly not be opened while the agent is running, they seem to self-close themselves again any time there's a content update And what's up with the complete radio silence and/or deflection of there even being any issue at all with how all this went down? What even is the roadmap for when this is planned to reach some kind of feature parity with v5? Will any changes be made regarding management of releases and QA etc involved in the future? Is all of this expected and just something the users should be prepared for with every future major version? Have any benchmarks been made before the release? Analysis of how token consumption is effected when switching from v5 to v7? Are any of these tests public? Did these KPIs even improve or worsen for the users? Is there even actual developers and science behind these release decisions in the first place or is all just a bunch of people with business degrees vibe-versioning software around here? Where is the transparency? Why no public statements that don't look like the PR department wrote them

The most recent update can be seen as a complete and utter betrayal of users.

Version 7 is utterly alien and broke the workflow unnecessarily. User has less control over the working process and the transparent price listing per provider/model is gone. I use Orchestrator a lot to work around the short context window of the lower coding agents, but now Orchestrator mode is depreciated. Switching between the modes will not fix the short context window issue and with it the context rot!

The version 7 is awful. It's too different from previous versions. Many customizable options are missing, and the interaction with the AI ​​model is much less transparent.

I worked very well with version 5, but the version 7 update is really bad, both in UI and UX.

I had tried version 7 when it was still in beta, and I'm baffled by the decision to fully release it in the current state.   Key features I relied on, like extensive config options for local, OpenAI API compatible models are gone. (No support for adjusting the context length? Really?) Displaying incoming LLM generated code diffs in the chat sidebar (after a click, for every file) is really inferior to using the vscode diff viewer. This completely breaks my workflow. I used to specifically recommend this extension to my peers that want to use LLM's as a pair programmer.  I'm also not impressed that importing my version 5 configs is broken - it set all the tools to auto approve (dangerous!) enabled auto formatting, and I had to reconfigure my OpenAI compatible provider.  I've never reviewed a vscode extension before - but the miss-management here needs to be called out. I will no longer be recommending this extension to my peers. After months of using KiloCode, I've switched to RooCode instead of downgrading to Kilo 5.12.0

Very inconvenient. Trust-breaking. Completely unreliable. I'm moving from it to its parent forks, either RooCode or Cline - now on my personal account, soon at larger scale.

Breaking update. All the functionality we used in Kilo lost in the version 7. Rolled back to 5.12. Requesting Kilo managers and devs to reconsider updating approach to incremental and non-breaking.

Why u not create a new version for anyone who want it and save old version for me and other users? U killed all progress! Bad, very bad decision. I switched to roocode and cline.

easy to use, VERY generous free tier and use with free models, allows you to connect pretty much any type of model provider in the settings, i havent had any speed issues or anything personally, and i personally think theyre the best out there currently in comparison to other similar extensions

Done.. completely done. I will be boycotting anything from your company from now on. Integrity is important to me, but clearly as part of your strategic/marketing decisions, you decided to go over to the dark side because you saw it as a path to greater wealth and usage. What you don't understand is, I was a supporter because of what I perceived was integrity. We all make mistakes and can recover from it, I hope you see the error in your ways and try to crawl back to the light again.

I used this every day, but now I’ve switched to Roocode and pay for my subscription through OpenRouter instead of you. This is unacceptable—a terrible update. You ignored these annoying bugs for months, and instead of fixing them, you just rolled out a completely new extension that doesn’t even cover half the features of the old one? Shame on you.

I've tried every mode this extension offers, and the result is always the same: it gets stuck in an endless cycle of asking the same questions over and over again. It doesn't matter which mode I'm in – the behavior is identical. The extension will ask a question, I provide an answer, and then it just asks the exact same question again. No progress, no action, no actual work being done. Just an infinite loop of pointless back-and-forth. I've spent far too much time trying to get this to actually do something useful, but it never moves past the interrogation phase. It seems completely incapable of executing any real task or generating any meaningful output. Save yourself the frustration – look for another extension that actually respects your time and delivers results instead of wasting it with repetitive, non-functional loops.

replacing an extremely convenient and functional design with something like this is highly questionable... not to mention that some functionality was simply removed. are you serious? unfortunately, i've had to switch service

Why is it that every time a company 'needs' to redesign a good product it turns out so much shittier then it was? If you can't offer feature parity with the old version; don't bother releasing it. I'll finish my remaining credits on V5 and will be moving on.

I knew it! Every time a product is good and people start to trust it, someone decides to ruin it.

Awful update on version 7. I switched to Roo instead. Here's what's missing and I need: - Code indexing has been completely removed. Critical. - Toggle for auto-approving MCP server use has been removed (MCP tools are now always auto-approved) - It's not possible to see or configure the tools of an installed MCP server anymore - Mermaid diagrams are no longer rendered inside the chat window - Auto compaction doesn't have a threshold setting anymore - Autocomplete is now locked to Kilo Gateway - The "File read auto-truncate threshold" setting (to limit how many lines of a file the agent reads by default) is gone - The "Maximum diagnostic messages" setting is gone - The "Include current time/cost in context" settings are gone - The "Terminal output/character limit" settings are gone - Support for image generation is gone - The beautiful task timeline that shows messages colorized by types is gone - Browser support for vision models is gone - Orchestrator is gone - All modes can spawn agents so they are all orchestrators for no reason - Support for profiles - Code review is gone. No, the current implementation is not even BY FAR the same - Checkpoints - Configuration for local LLMs - Command copy buttons in timeline Did I mention indexing was removed? For anyone looking for local indexing of your own data look elsewhere. This is no longer possible in Kilo. Roo is your only option now.

Tried a lot of different code buddies and Kilo-code was my favorite. Sure it had some quarks - but usability, efficiency, MCP integration, and multi-model (kilo router) program was really appealing. And the app was maturing nicely for a while. I was excited for the latest 'paralelling' update until I installed it, nothing worked, I had to redo all my settings, and even then - it lost a lot of useful functionality and tanked efficiency. Now has become a credit gobbling gremlin. I was a kilo-plus subscriber. Canceling that and rolling back to previous version while I look for alternatives. Giving 2 stars. +1 for what it used to be (and the fact that you can still roll back)

I used this tool despite all the bugs and issues because there just aren’t any better alternatives out there. I even built my entire workflow around it. Now, I really regret investing so much time into Kilo Code. I definitely won’t be renewing my subscription for their AI endpoint. The developers released a new version out of nowhere and stripped away several features. The ones that are left are even more unstable than before. To make matters worse, they clearly didn't bother testing the migration process because every single one of my settings is gone. I’m not even going to bother restoring from a backup. I’d rather look for something else because you just don’t treat users this way. Edit: Probably the most serious change is that they dropped the orchestrator and its per-task reports. This was the only mode that allowed for building complex apps while keeping the context intact. Now, it's impossible to use Kilo Code if you need a truly well-organized workflow. The cherry on top is that codebase indexing has been removed! It doesn't matter if it's temporary or permanent. Basically, the developers expect me to pay three times more for increased token usage because the LLM now has to figure out every task from scratch without any codebase context. All these changes feel like a huge slap in the face to developers who want to do high-quality, efficient work.

The new version missing a lot of features and hard to use. Where is the review mode? Where the quota viewer for Codex plan? Where is the full code indexing? Where are the checkpoints? Why all modes are the same now? And that's just what i noticed missing in the first hour of use. Edited: - /local-review is not the same as review mode and i can't add rules to it - the first thing i did was enable checkpoints yet i can't see it

I wouldn't say that the new version is much worse, but it's a clear step back. Many features have simply disappeared, there are incomprehensible changes in agents, and in general there is a feeling that the developers movement is no longer on the same path as the users, but let's see what happens next, there are enough tools on the market to not miss a kilocode.

New version 7 lacks a lot of stuff which made Kilo Code outstanding among other extensions: UI for auto-confirm commands, file reading settings, UI for local LLM provider, non-bloated model selection, context progress graph, copy buttons in chat, UI for skills, commands and MCPs in-chat. Now all of this is gone and for what? Just now need to find another alternative.

Kilo code was pretty enough until 7+. It even couldn't completely migrate settings from previous installed version! Already rolled back to 5.12

Kilo code was great until 7+ version. It even couldnt migrate settings from 5.12! WTF?!

The new version is terrible, I've already rolled back to the previous version.

The extension is clearly unfinished: the login doesn't work, migration doesn't work, and everything else is irrelevant as a result. I rolled back to the previous version. (https://disk.yandex.ru/i/CfAf0kLUbNbWkQ)

For two weeks, I tested various plugins for vibe coding with a local model. It felt like I'd finally found the perfect one—everything was visualized, it displayed a beautiful diagram, and it was one of the few that let me modify and preview the system prompt. Then today, after hours of meticulous tweaking—adjusting prompts, rules, and LM Studio settings—I was on the verge of finishing my project. Suddenly, an update rolled out… and everything broke. I had to reconnect the provider all over again. Now the settings include random Linux commands that don't even exist on Windows. Previously, when confirming file changes, the plugin showed a clean diff directly in VS Code, so I could clearly see which lines were added or removed. Now it dumps the entire file—hundreds of lines—right into the chat, making it impossible to tell what changed or what to compare against. I'd roll back to the previous version, but where can I even download it? They broke everything. I'm devastated…

Well. Comparing to almost any other Agent this one is.. Heavy Thinker. Script 300 rows, thinking 1000 rows 4 times for change of 1 row. Never use this garbage, its just token eater, also cant configure thinking.

I love Kilo Code for its rich featureset, but this has been completely undone in the newest version (7.1.8 and onwards) where the extension has basically become an advertising platform for Kilo's own paid models. JUST INSTALL *5.12.0* NOT THE NEWEST and you should be fine for now.

The user experience is very poor, there is not even a conversation rollback function.

Horrible. If I pause, make a new question, the LLM using kilo code ignores the question and continue working, always ignore my question and this behavior only happening in Kilo Code. I already report that and I'm offended about the kilo code team. The worst Ai Coding Agent. PS. i'm using Opus 4.6

Kilo Code is very nice.

Had a weeks of work with this Extension in VS Code and Cursor. Both pre-release and release versions have numerous bugs, and Kilo's support just suggests to switch versions, none of which is working. I spent more time trying to make things work and fixing bugs than actually developing something. I spent money, and most importantly - time. Avoid this until they make something stable

Tested it, it took 10 minutes to finish a simple fix, it failed to make the fix, and then it decided that that the free version has ended. So, what exactly did I get from this?

Норм. Ниразу не зависала и не падала

Subscription Trap: Unauthorized Auto-Charge and Blocks Card Removal Warning to all users: I tested Kilo Code and paid for credits, but the platform has a predatory billing system. Blocked Card Removal: The UI does not allow you to delete your credit card details while a subscription is active. This is a massive security and privacy concern. Unauthorized Auto-Renewal: I was automatically charged $19.00 for a renewal I did not authorize or want. Refusal to Refund: Despite contacting support immediately, they refused a refund, claiming "Terms of Service," even though their interface makes it intentionally difficult to manage your own billing. This is a classic "subscription trap" with a very bad user experience (UX). They hold your credit card information hostage until you figure out how to navigate their cancellation process. Avoid this service if you value control over your own financial data.

Auto edit in file - that's why deleted this peace of bug code. And a lot of bugs. No reason to use it. GML 5 free it was only the one reason.

You completely killed the best plugin of its kind. “New” plugin is just a skeleton without functionality. Terminal? TBD. Agents? TBD. Providers? Absolutely not working. Old configurations lost. And where the hack is JetBrains plugin??? You killed the radio star ⭐️. VS Code plugin is just for quick tests for me. You did the mess and even didn't bother to inform users. BTW. I’m your blog subscriber! (I was…)

Love it; it works very well for me and my workflows. The only considerations are that they should show the token price, as in the CLI version and also the profiles should work in all the modes, so if I change my profile (for example, I have my own keys for OpenRouter and other models), it should change for every mode. It’s not good that I change a profile for coding and then, in Ask, it’s a different profile

Probably the best Cline-style AI coding extension for VS Code. Great configuration options, custom modes (make your own AI agents), works like a charm. Well done!

May be the best AI Assistant so far, with great features, rapidly evolving on weekly bases. Feature request for the Developer Experience in enterprise environments: at the proper time, think about Visual Studio extension too. Good luck!

Looks like they introduce erros on purpose to use Ollama local models, because cloud ones works

The application consumes more tokens than it should.

Realy a great ai development extension. I use it with context_portal and qdrant. When starting a new project you should always take the time to meticulously create a consistent memory bank. It will later help you to keep track of design decissions you made. Read the docs ;-)

Reading a PDF file consumed all of my tokens.

Actually, Kilo Code is top-notch and could lift your code quality and performance way higher than you would experience when using an LLM without an agent. The benchmarks speak for themselves. Fully recommended, especially for full-stack developers / "no coder”.

i got banned for no reason :[

I want to like this. I really do. But it's so buggy, and things are breaking constantly. Every week, one model breaks that worked the previous week. The settings page is constantly in a buggy state--this week's bug is that creating a new provider profile actually doesn't create a new one, they're both tied to the same config behind the scenes, so if you update one, they all get updated. The GitHub issues are backed up out of control. I really don't want to go back to Cursor, but I'm spending more time fighting this extension than I am actually using it.

Every time a project is opened in the latest VS Code version, the Kilo extension's view automatically pops up in the left sidebar. Please provide an option to disable this auto-reveal feature.

Thankfully, I only tried $10; it was indeed unusable. And when using Claude Sonnet 4.5, generating a 1000-line document cost nearly $1? I promptly gave up.

I don't know why the number of tokens consumed is beyond imagination!!!

Very good

If you are using your own API provider, the Autocomplete feature will not work. Waste of time

Best so far

Quickly becoming my favorite open source IDE agent. Cursor, Replit, etc. are great , but I do not like getting used to a proprietary toolchain. In the long run open source and choice wins in my experience. It is impressive how fast the Kilo Code team merges the best of other popular open source agents (Cline, Roo) , keep up with frontier APIs (e.g. first to integrate Grok Code Fast and Supernova) as well as MCP tools.

what would take you under 10 to 20 minutes on codex / claude code will take you a day on kilo code

Currently the best out of Cline forks out there. They added a lot of useful features providing better control over the configuration details. The only problem is that after a long run I often get it crashing - the whole side panel becomes gray and requires VSCode restart to fix.

crashes my editor

slow and laggy

J’utilise Kilo dans le cadre d’une application que je développe pour automatiser la gestion de factures. Mon workflow : dépôt de PDF → découpe intelligente → extraction en JSON structuré → intégration directe dans Airtable/Glide. Kilo se révèle très fiable pour uniformiser des documents hétérogènes (factures, devis, bons de commande) et produire un schéma strict, toujours complet, sans erreurs de format. L’ergonomie est claire, la stabilité au rendez-vous, et la possibilité de personnaliser les agents me permet d’adapter finement le traitement à mes besoins métiers. Comparé à d’autres solutions (Cline, Roo), Kilo offre une meilleure expérience et une productivité accrue. Pour tout projet nécessitant de transformer des documents en données exploitables, c’est une brique essentielle.

The level of control you can have over your agents is great. I love that it comes with 5 modes out of the box.

I've been using Kilo Code for a while. It has a lot more usability and is more stable, and actually fixed numerous issues I've had in the past. No other AI agents come close.

I have tried many AI coding tools, and I really like Kilo Code because of its agentic editor. It combines planning, coding, and debugging in a seamless way, and the multi-mode system makes development faster and more structured. While it can be pricier than some alternatives, its flexibility, extensibility, and productivity boost make it one of the best AI coding assistants I’ve used.

Excellent AI Agent for Developers I’ve been using Kilo Code AI Agent for a while now, and I’m really impressed. The module keeps getting better with every update — faster, more accurate, and genuinely helpful in day-to-day coding tasks. It integrates smoothly with my workflow in Visual Studio and feels like a natural extension of the IDE. If you’re looking for an AI assistant that actually evolves and improves over time, I highly recommend giving this one a try.

Best AI Agent I've used. Kilo code has better UX than both cline and Roo code and is a lot more stable during my use.

Love this product. I have been using normal llm so far to do the AI assisted coding to speed up development. This workflow is good and it is similar to the workflow i have adopted using standard LLM but without the overhead of myself doing the steps manually and it is good at maintaining consistency in the development process

Hi Kilo team 👋 I just started testing Kilo Code and noticed that I only received **$5 in free credits**, while other users mentioned getting $20 or even $25 when they signed up earlier. Because $5 runs out pretty quickly (especially when testing Architect/Debug modes on real projects), it feels a bit limiting for properly evaluating the tool. 👉 A couple of suggestions that might help: - Keep the free credit amount consistent for all new users. - Alternatively, allow more free usage on smaller/cheaper models so people can explore the workflow before upgrading. - Add clear documentation about the current free credit policy, so expectations are set up front. That said, I really like the concept behind Kilo Code—especially the Architect → Code → Debug flow—and would love to keep testing it further. Thanks for building this tool! 🚀

Asking for credits in the middle of a task and NOT even using their AI!!

I like the ability to create your own custom agents and rules. gets job done the way you want it.

Would be awesome if Auto button too is there instead of letting users select the model

Vraiment très impressionnant et très configurable : On peut travailler en direct avec ses propres clés d'API ou un LLM en local. Naturellement, les serveurs de Kilo permettent d'accéder à tous les grands modèles. Une extension à tester d'urgence pour tous les développeurs. Really impressive and highly configurable: You can work live with your own API keys or a local LLM. Naturally, Kilo's servers provide access to all major models. An extension that all developers should try out immediately.

Это худшее что со мной было. Были агенты которые предлагали глупости, не видели контекста, но чтобы он сам предложил изменения и сломал абсолютно всё чего коснулся, такое впервые. Худший агент, не понимаю откуда высокие рейтинги.

Finally, after a lot of trial and error with agentic tools, I'm getting into some impressive vibe coding tools.

Kilo Code is smarter than GitHub Copilot extension for VS Code and does a really good job.

I'm extremely impressed with KiloCode.ai. As a developer, I'm always looking for ways to streamline my work, and this tool has exceeded my expectations. The AI code completion is incredibly smart, understanding context and suggesting high-quality code snippets. The interface is clean, and it integrates seamlessly into my workflow. It's a must-have for anyone looking to boost their productivity.

I have been using Kilo Code in Cursor the last couple weeks and have found it very beneficial. I like seeing the flow of the conversation and cost at the top of the window. The ability to switch profiles easily so I can change which API keys are billed has been very helpful to my workflow.

Great extension! Love that it has all the features from both Roo and cline. it also incorporate features from other ides like the workflow feature from windsurf. Great app and great team behind it

I like KiloCode it mix both roo code and cline in one and adding some unique features. Thank you the effort