Cursor Reviews 223

By every version update, they make your work experience worse and worse, and your costs skyrocket.

In March 2025, Cursor worked well. It was fast and convenient. They had a "Manual" mode, where responses and code edits applied immediately. And an "Agentic" mode, which was interesting (e.g., could execute commands) but I didn't use it. Not needed for my work.

After a while, they removed "Manual" mode. The "Agent" mode was terrible and buggy. It didn't support our favorite models, which weren't trained for agentic use. The agentic mode with agentic models more than tripled the cost of our edit tasks, since every tool call was an extra model call.

Thankfully, we could use custom modes and create our "Legacy Manual" mode that didn't use tool calls and worked with custom models. It was never the same again. The system prompt changed. The model believed it was in agentic mode even when disabled, but it was at least usable.

Then in the recent 2.1 update, they removed custom modes too. The alternative "custom commands" feature doesn't substitute the lost features. Now Cursor is totally crippled for serious work, e.g., single-file code edits, CUDA kernels, etc. It's slow, costly, and error-prone to use "Agent" mode for such tasks.

Cursor is now a tool for vibecoder kiddos who want to one-shot scaffold their new TypeScript webapp. The Cursor team aren't serious people. Don't trust them.
I didn't even go into their drama about pricing changes mid-year.

Find a serious alternative product. GitHub Copilot in VS Code is good for such non-agentic, assisted coding tasks. Windsurf is almost at this level but can be pricey too. Even the JetBrains All Products pack contains a bonus AI assistant subscription.

Or if you want to just vibecode your latest AI image generator SaaS slop, use Cursor.

Terrific experience trying to get the student discount. An odyssey through stupidly useless SheerID verification and their support team which has no will/power to provide actual support. Will switch to Windsurf!

I have tried, paid for many ai ides previously, was happy with one of them because a year ago when I tried cursor it was so complicated for a non code expert like me. A month ago I have given Cursor another shot, it was like I was missing a lot! I can confirm and God Witness my testimonial that Cursor is the best AI Powered IDE in The world. Fast, Accurate and perfect.

Extremely Bad Token Management.

This could be a good tool, i was expecting a lot from it.

You can't be anxious at every single prompt to have 1m+ token spent.

This is not the best tool.

I am incredibly frustrated with the experience I've had with Cursor.

I signed up and was approved for their student plan back in May with my Open University email (a valid .ac.uk address). I used the service without any issues for over four months.

Around the end of September, my student access was suddenly revoked. I received zero notification about this. No email to my student account, no email to my personal account, no warning that I needed to re-verify. The plan was just cancelled.

I then spent over two weeks just trying to get a straight answer from support. I was initially blocked by their AI agent, "Sam," which was useless. When I finally reached a human (Cheska), I was put into an exhausting support loop.

Despite me sending screenshots and explaining the situation repeatedly, I was just sent the same canned responses and directed to a verification link that did not work for my situation. I asked to speak to a manager, and my request was completely ignored.

Only after two weeks of this, and after I had to push back multiple times, was I finally told the real reason: their policy had changed, and they now only accept .edu email addresses.

This is an awful way to treat your users.

Why was I approved in the first place if .ac.uk was ineligible?

Why was my existing, approved account cancelled without a single word of warning?

Why did it take two weeks of circular nonsense and an ignored request for a manager to get this simple, final answer?

The policy change is one thing, but the complete lack of communication and the time-wasting, abysmal support is unacceptable. Be warned if you are a non-US student.

Dear Support Team,

The payment for using your service — specifically for token usage — does not correlate with the actual use of the agent. It’s unclear how usage is connected to billing. Please review my recent transactions, and you will understand the issue.

Until you fix the problem with billing limits, I will not be using your service. At the moment, I’m charged more when fewer tokens are used, and charged less when more tokens are used. I have set monthly spending limits, but they are not being applied correctly.

This is an issue on your side, and as a result, you are losing a paying customer. If you resolve this problem, I may consider returning to your service. However, as long as random amounts are being deducted for the same agent and there is no control over my subscription, I refuse to continue.

I had an active spending limit, but it was ignored. Until these issues are fixed, I am not going to use your service any further.

Best regards,
Your best customer

This s*** company has HORRIBLE transparency when it comes to their pricing plans. CURSOR - YOU GUYS better figure it out before you start fires you can't put out.

What initially seemed like a good deal turned into a nightmare.

I really like it for code exploration, and so on, but for coding it is completely useless - not that they did their job, but these AI tools make coding really annoying.

Also, the IDE got recently super slow, and as someone who uses vim regularly, this is absolutely unacceptable.

So while it is it definitely interesting and worth trying and has its target audence - I am not one of them.

The system is becoming increasingly inefficient and limited. It fails to read files, analyse existing code, and provide coherent responses. The internal browser is completely useless; in fact, it only leads to errors. It is impossible to perform a real analysis without having tons of code and files written without the slightest logic or evaluation. Interacting with the agent becomes frustrating, as it is impossible to give it the slightest indication. Instead of supporting development, it slows everything down and introduces inconsistencies and errors.
I also get the impression that the agent aims to do as little as possible, suggesting that you skip certain parts (to run complete tests, you have to ask at least five times) or avoid certain steps.
It is not worth the expense, honestly.

Cursor used to be good, now its lacking. I noticed that the models are getting dumber. It seems like they are trying to cut API costs by either giving less context, or secretly calling a cheaper model instead of the model you pay them for. And for this reason, I will no longer be using their services.

Unclear billing practices, they scam you into forgoing the 30 day trial, then once you do sign up, there is ZERO support. Emails to support go completely unanswered.

Our company has been using this service for quite some time, but the recent change in management UI for company accounts has been a complete disaster.

After the update, several of our developers were suddenly downgraded from the paid plan to the free plan, even though our organization is still actively paying for the service. To make matters worse, the admin dashboard no longer provides any way to manage users or make plan changes — it’s as if the control panel was simply removed.

We reached out to support over a week ago, and the experience has been extremely frustrating. Responses are slow, unhelpful, and repetitive, with no real progress or resolution. Essentially, we’re paying for a service we can’t even access properly, and support keeps promising to “escalate” the issue without any visible results.

This isn’t just a poor user experience — it’s unacceptable for a paid business service. Until this is resolved and the management tools are restored, I can’t recommend this company to anyone relying on it for professional use.

$200 Pro plan, exhausted in 5 days. $50+ additional tokens = $200 per week, to use AI Cloud Models that I already own?

DeepSeek, $10 a year, used only $5.00 worth in 12 months.

Perplexity Sonar models, $20, only used $5.00 in two months of daily usage.

Issues:
Cursor's auto switching models in mid-dev = Broken code, no memory.

Cursors broken chat panel scrolling bug, never fixed. Forces the user to repeatedly scroll to the bottom of the thread. Scrolling options do not work.

Broken Context tags: Perpetually removes Docs, Files, Rules from chat = More broken code amd bugs.

Cursor's auto switching from Agent mode to Chat mode = wasted tokens, increased context usage, flooded chat threads.

No chat save to Markdown feature = poor performance of chat context awareness.

Blocked from adding your own AI Cloud Model API keys: Gemini DeepSeek, Perplexity, Claude, etc = More proxy middleman reseller fees and security risks.

Total disregard of app settings to NOT USE CERTAIN AI MODELS. Disabling Claude, doesn't work. They force Claude on you and auto switch without consent or obeying rules. Because, well, they keep more money from the premium AI model tokens.

The age of middleman reseller proxies exploiting people is gonna die off soon. A wrapper, reselling a wrapper, reselling a wrapper, just to bill people more for AI Models they already own a cloud account with.

Just BYOK, i recommend using Zed IDE, or Theia IDE. App that ACTUALLY WORK. You'll save THOUSANDS OF DOLLARS AND LESS HEADACHES.

I subscribed to Cursor Pro on October 18, 2025, hoping to use it for converting a simple HTML page into a Next.js project. Unfortunately, the experience was extremely disappointing.

After using Cursor for just one day, my entire project folder disappeared when I closed the app — all files and code were lost. I tried creating a new project afterward, but the functionality didn’t work as expected, and I couldn’t generate or reuse even a single line of code.

When I requested a refund, I was charged $10.51 but was refunded only $9.49, which doesn’t even cover the full amount — not to mention the time and work I lost in the process.

It’s very frustrating to lose all your code because of an IDE, especially one marketed for AI-assisted productivity. I cannot recommend Cursor based on my experience. Their refund process and data reliability both need major improvement.

I'm surprised by the poor reviews. It works well in my experience, not a replacement for coding, but takes out much of the drudge.

I want to warn other developers before they lose money like I did.
Cursor AI’s billing system appears to have serious issues with token usage calculation.

Here’s what happened:
For identical usage of around 700K–785K tokens, I was charged ~$0.60 multiple times — which is normal — but suddenly $6–$7 for the exact same token amount in other sessions.

I’ve attached screenshots showing multiple examples of this inconsistency. In just one day, Cursor charged me over $210, which is completely impossible based on their published token rates.

When I contacted support, I received no proper explanation. If their pricing logic was accurate, every 785K-token session should cost roughly the same — not 10× more at random.

This isn’t a small billing glitch — it looks like a systematic overcharge problem that can silently drain your balance without reason.

I strongly recommend everyone to double-check your Cursor AI usage logs and invoices carefully. Something is clearly wrong with how token costs are calculated.

Cursor AI team, please fix your billing logic and issue refunds to affected users.

I renewed my Cursor Pro subscription after a failed payment, but I was very disappointed to see that my billing period didn’t restart from the day I paid. Instead, it was set from the previous billing date (September 11), so I ended up paying $20 just to get 2 days of access until October 11.

This feels really unfair. I expected my new subscription to start from the day I paid, not from a past date. I’ve contacted support to fix this issue, but I believe the billing system should not work like this in the first place.

I hope Cursor resolves this quickly, because I actually like the product — but this kind of billing practice is frustrating.

I am extremely disappointed with my experience using Cursor.

I upgraded twice, first to the Pro plan for $20 and then to the Pro Plus plan for $37.63, yet I could not use any of the models from day one. While everyone else seemed to be using Cursor without issues, I was completely locked out for my entire subscription period.

After contacting support, Kemp from Cursor admitted that the issue was on their side and caused by a service outage from their model provider. He apologized and assured me that a full refund would be processed immediately.

Weeks later, I have still not received my refund. Instead, another representative, Neil, denied the refund and claimed that I had used over $216 worth of value, which makes absolutely no sense considering I could not access the platform properly in the first place.

I have spent $57.63, received no value, and been given completely contradictory answers by the support team. The lack of consistency, accountability, and transparency has completely destroyed my trust in this service.

Cursor may have potential, but this kind of customer treatment is unacceptable. Until they fix their refund process and communication, I cannot recommend Cursor to anyone.

I subscribed to Cursor AI Pro, made the payment through their dashboard, and even received a receipt. Shockingly, I later got an email saying my payment was not processed. Till today, my account still shows only the trial version.

This is beyond frustrating. I have already shared proof of payment, but their team has not provided a single proper response or resolution. Their dashboard, email system, and IDE are completely out of sync — it feels like an unprofessional and careless setup.

Cursor AI clearly does not care about its customers. It looks like their only goal is to grab money once and then abandon the customer. This is the worst kind of service you can expect from a so-called “AI company.”

If you are thinking about paying for Cursor AI, don’t make the same mistake I did. They deliver zero value, zero support, and a horrible customer experience.

Final words: Cursor AI is unreliable, unprofessional, and not worth even a single dollar.