Kilocode Bewertungen 

going round in circles and messing up code that worked before. stay away from this shit.

Very good AI coding agent. Also, good support who is eager to address any problems. I can recommend their MiniMax M3 plan. Just as good for agentic coding as Sonnet 4.6 or GPT 5.5 but at a tenth of the cost.

Kilo code with z.ai is a winner.
Once you set up the mcp servers (see z.ai site or ask Kilo code for help), then it works really well, reads debugging output images, figures issues with it's own code and solved problems well. Beats Cline,Continue and Roo for me. It's the one for z.ai!

I used Kilo Code via the VS Code extension and it was a frustrating experience. It’s absolutely not worth your money. I wanted to try different models, but it frequently got stuck on “Considering the next step” and often took a very long time to respond. In practice, you can’t reliably get work done with Kilo Code. Completely unreliable. Save your money. Thank me later.

As a senior dev who's tried plenty of AI coding setups, Kilo (Kilocode) stands out for keeping things practical and open.

I appreciate how the team focuses on lowering barriers. **KiloClaw** offers one-click hosted OpenClaw with full model access, skipping local setup headaches. **Gas Town by Kilo** provides a managed version of Steve Yegge’s multi-agent orchestrator, with an open beta available.

It's not the most refined tool on every front yet—some edges are still evolving—but it's genuinely community-oriented, transparent, and growing through real feedback. With everything open-source on GitHub and active development, it feels like a tool that's improving alongside the people using it.

If you're looking for something less locked-in and more straightforward, it's worth a try at kilo.ai

This is the worst AI IDE I’ve used and not worth trying.

I’ve been very patient and have given it multiple attempts, but I’ve never achieved good results—even while on a paid subscription.

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.

The structured approach to task execution is initially impressive. That said, I've encountered some severe bugs.

Context confusion: In one case, I asked it to explain its orchestration mode. Instead of answering, it latched onto an example from its own explanation and started implementing it—suddenly announcing "I'll design a database schema for user authentication." When I tried to course-correct, it stayed on track for one response before reverting right back to the database schema. Completely unrelated to my original question.

Compaction issues: Mid-way through an architected task, it suddenly referenced a question I'd asked 20 minutes earlier, treating it as the current request. This kind of context drift can derail longer workflows.

Provider configuration: There's a bug where the configured provider isn't consistently respected. I had it set to use one provider, but it silently used another—resulting in unexpected charges to a different account. Thank you kilocode for burning up all my credits on an API plan instead of using my Team plan.

you cant be more useless than this app wasting api tokens.

They deserve it for draining my deepseek specialle api with only errors like wtf is this product? I get is free, my api was not so why release products that dont work with external providers? thats the same as lying, scamming or running malware.

But is clear, we live in the mediocre devs era.

Honestly, if the UI was better developed for the CLI version of KILO I would probably be inclined to rate this a 5/5. It's an outstanding addition to VSC or Cursor and it almost always has a featured free use model for testing. The amount of work I accomplished while using Fast Grok's free use period was quite astonishing. And while the CLI isn't bug free I admit I still use it regularly. It's a great addition to any workflow. Be it coding, design, or pen-testing you should find KILO helpful. I highly recommend trying it out. Like any other AI model, I will add that you should approach KILO as a tool in the belt. It is not a one stop solution to problems or projects.

I purchased $100 in credits and received an additional $20 promotional credit. After my project was canceled, I asked for a refund. I had only run a few tests that used about $4.

Refund experience:
Support declined the refund, pointing to the user agreement and the fact that any usage disqualifies refunds. Practically speaking, my $100 purchase was never meaningfully used, yet the policy left no flexibility.

Product experience (VS Code extension):
To at least use the remaining balance, I built a small web application that calls an external API and displays the results in a simple UI. Despite providing a structured spec (SSD/speckit), I repeatedly encountered:

Frequent rate-limit and unknown error messages

5–15 minute waits per task

High token consumption relative to progress

These interruptions made iterative development frustrating and expensive.

What worked:

The initial credit bonus was a nice idea in theory.

What didn’t:

Rigid refund policy with no accommodation for minimal use and canceled projects

Unreliable VS Code extension with frequent errors

Slow task turnaround and inefficient token usage

Verdict:
Given the combination of inflexible refund terms and unstable tooling, I cannot recommend Kilo Code. In my experience, established AI coding platforms offer more reliability and better value.

For first ,advertising kilo code for devs is fail , maybe not if you use it for creating documentation or simple html , working on mvc failed so bad it is painfull . It stucks in loops , spends so much tokens that it hurts but resuls are bad . It cant complete one task without looping or just loading for century . In my case php mvc was issue for kilo code , it cant handle it at all ,only thing that i can say it works as expected is debugg mode , works very good

Unlike their advertisements Kilo Code is 10x more expensive than Github Copilot in VS Code. It burns through credit like nothing. And it makes exactly the same mistakes, just for more money.

average, not very good money wise if you are working on real software engineering complex tasks.

I want to sincerely thank Kilo Code — they helped me tremendously as someone with little to no experience in game development or app programming in general.

I was honored to be the first to give them this evaluation, and I’m genuinely proud of that. They encouraged me from the very beginning and generously offered me two trial opportunities at their own expense.

They truly delivered — with passion, creativity, and real success.

I highly, highly recommend them to anyone considering a tech project. But remember: to succeed, you need to use Kilo Code wisely.

Thank you, Kilo Code — I’m grateful and excited for what’s next.