I use Cursor for my everyday programming and coding sections, and it helps a lot on my projects. It's very simple to use with a great learning curve, and it's similar to other famous IDEs like VSCode. The interface is straightforward, and I can explore deeper to get the most out of Cursor. It's fantastic for switching between different models and modes, with everything very integrated and simple to use. The fact that it's well integrated with AI and offers great subscription values that are cost-effective for using AI on projects led my team to switch from VSCode to Cursor. Additionally, the initial setup was straightforward, with self-explanatory buttons that made it easy to sync my other projects. I totally recommend Cursor and have already done so multiple times. Review collected by and hosted on G2.com.

Maybe it is not very well explained which models are best for what you're doing, because there is quite a lot of models, but I don't really know which one is the best for what I'm doing. Maybe just a little info about what the model is best for, like this model is best for coding generation, this other model is best for design and stuff. Review collected by and hosted on G2.com.

I really like to be able to change from agent to planner, to debugger, tu multitasker, to ask mode,, this make is easy and clear, I like composer 2.5 becoming more capable AI with good performance , i like the overall IDE, easier for me to work than from a terminal, and i like that there is a lot a easy accessible setting, I find the pricing very expensive but i am learning to work with it more to use the right tool/model at the right time. the intergration are easy to use, just ask and crusor installes them, never really contacted support except one for a billig question, got a response right away, good service . Review collected by and hosted on G2.com.

Other than sometimes composer make mistakes ,, that i have to fix with opus.. i like all, need to select what to ask composer Review collected by and hosted on G2.com.

The speed of the Composer models & the possibility to run prompts in ||.

The pricing is quite fair too. Regarding the UI I like the leading placement of the chat conversations and the management of || chats.

Integrations - I like the possibility to let Composer do some extra tasks directly with the DB or the servers.

Onboarding - it was easy to set everything on my own or with some tips from the community.

AI features - the fact that Composer helps in configuring the IDE & resolving issues is outstanding - this is the way apps should go. Review collected by and hosted on G2.com.

My main issue so far was that the .git folder messes up with iCloud sync daemon. Some files disappear & reappear in my project. Review collected by and hosted on G2.com.

It empowers us to accomplish more, better, and faster. The agents can complete tasks and features that used to take me a week in just a few minutes. Every user suddenly has a team of developers working alongside them, moving faster than I ever could on my own. Review collected by and hosted on G2.com.

Honestly, I don’t have many bad things to say. Every now and then there are connectivity issues with the model providers (e.g., ChatGPT or Claude), but that feels like nitpicking. On larger projects, it can also lose track of the project requirements because of the limited context window, although this seems to be improving as the models themselves get better. Review collected by and hosted on G2.com.

It keeps your feet on the ground; don't just complete tasks recklessly. Review what you do and, if your idea of a function or code isn't good, it suggests better options. Moreover, it shows you more viable and clear paths to reach the result.

You can decide to stick with your initial idea, but generally its suggestions end up being better. Review collected by and hosted on G2.com.

I think that, so far, what I like the least is that, in terms of the UI, it is not usually as creative as Gemini or Copilot. Even with the superior backend that Cursor offers, sometimes I have to resort to other AIs to improve the front end. But, honestly, this is not a big problem. Sometimes the prompt itself wasn't adequate; if I improve it with another AI and then pass it to Cursor, it usually gets resolved. Review collected by and hosted on G2.com.

What I appreciate most about Cursor is how naturally the AI assistance integrates into my coding workflow. It feels noticeably faster and more context-aware than traditional autocomplete tools, especially when I’m in the middle of a larger change and need suggestions that truly align with what I’m working on. The vast set of agents provided by cursor is also remarkable. Review collected by and hosted on G2.com.

The first issue for me is the expense. At first, it was priced per request, but now they’ve upgraded it to be token- and context-based. One thing I dislike about Cursor is that the AI suggestions can sometimes be inconsistent or overly aggressive, especially when I’m working on complex logic or doing large refactoring tasks. Review collected by and hosted on G2.com.

I like the IDE experience in Cursor, especially compared with other tools like Claude Code and Pi Agent. I can easily review code changes just by opening the files, and it’s straightforward to revert changes when needed. Sometimes I only need a bit of fine-tuning on the current edits, and Cursor makes it easy to build on top of the work that’s already been done. I also don’t have to explain things to the AI model again and again—I can just have it go and make the change directly in the file.

The file history feature is also helpful when I need to revert a specific file. On top of that, it’s easy to configure MCP servers, and I like that there’s a GUI-based configuration option as well.

However the Auto mode is less intelligent and using premium models is expensive. Review collected by and hosted on G2.com.

The Pricing and less intelligent in house models. Review collected by and hosted on G2.com.

What I like best about Cursor is that it feels like an actual coding tool built for developers instead of just an AI chatbot inside an editor. The biggest difference for me is the tab completion experience. Unlike CLI agents that generate code somewhat blindly and make broad changes without much visibility, Cursor keeps you in control. You can see suggestions inline as you code, accept them piece by piece, and guide the implementation naturally instead of handing over the entire task.

The contextual autocomplete is incredibly good for real development work. It understands the surrounding files, existing patterns, variable names, and architecture, so the suggestions often feel like something I would have written myself. It saves a huge amount of time on repetitive coding, refactoring, boilerplate generation, and debugging.

I also really like how flexible the platform is. Being able to choose different AI models or connect your own API keys is a huge advantage. Different models work better for different tasks, and Cursor gives you the freedom to optimize for speed, quality, or cost depending on the workflow. That flexibility makes it much more practical than tools that lock you into a single model.

From a performance standpoint, Cursor feels fast and responsive even on larger projects. The inline suggestions appear quickly enough that they don’t interrupt flow, and the editor itself still feels lightweight because it builds on the VS Code ecosystem. I was able to keep my existing extensions and setup without needing to rebuild my workflow.

The onboarding experience was also smooth. Since the interface is familiar, it only took a few hours to fully integrate Cursor into my daily workflow. The documentation and setup around models and API keys are straightforward, especially for developers already comfortable working with AI tools.

In terms of ROI, Cursor has easily paid for itself through time saved. Tasks like writing repetitive code, debugging, refactoring, and understanding unfamiliar codebases are significantly faster now. It reduces a lot of the small friction points that normally slow development down throughout the day.

Overall, Cursor improves productivity without taking control away from the developer. It enhances the coding experience instead of trying to automate everything blindly, and that balance is what makes it genuinely valuable. Review collected by and hosted on G2.com.

One thing Cursor could improve is consistency across different AI models and workflows. While the flexibility to choose models and connect your own API keys is a huge advantage, the quality of responses and edits can vary depending on the provider. Sometimes one model handles context perfectly while another loses track of project structure or makes overly broad changes. Smarter defaults and better guidance around model selection would make the experience smoother.

The AI can also occasionally become too aggressive with edits when using larger code actions. In some cases it rewrites more code than necessary or changes files outside the intended scope. It is still better than fully autonomous CLI agents because Cursor keeps the developer in control through inline suggestions and reviewable edits, but the AI can still feel overconfident at times.

Performance is generally very good, but there are occasional stability issues around extensions and AI panels. I have run into situations where extensions like Codex fail to load properly, get stuck on loading screens, or disappear entirely until Cursor is fully restarted. This does not happen constantly, but when it does, it interrupts workflow and can be frustrating during longer coding sessions. There are multiple similar reports from other users around extension loading and chat panel reliability as well. ([GitHub][1])

On larger repositories, indexing and context retrieval can also become inconsistent. Sometimes suggestions lose relevance or the AI misses nearby context that feels like it should have been included automatically.

Pricing can also become a consideration for heavier users, especially when using premium models with API-based billing. The flexibility is great, but costs can increase quickly if you use high-end reasoning models throughout the day. Better usage analytics and optimization suggestions inside the product would help users manage spend more effectively.

From a UI perspective, the core editor experience is excellent because it builds on VS Code, but some AI-related controls and settings still feel scattered across multiple menus. Features like indexing behavior, model routing, permissions, and agent configuration could be organized more clearly.

Support and onboarding are solid overall, but newer users may not immediately understand the workflows that make Cursor truly powerful. The tool becomes significantly better once you learn how to guide the AI properly, combine chat with inline editing, and use different models strategically. More advanced onboarding examples and workflow templates would help shorten that learning curve.

Overall, most of the downsides are related to polish, reliability, and predictability rather than the core product itself. Cursor is already one of the best AI coding environments available, but improving extension stability, large-project consistency, and cost transparency would make it even stronger.

https://github.com/openai/codex/issues/17290 Review collected by and hosted on G2.com.

It's the perfect tool where I get the best of all worlds - from being able to use different models for different functions, sync with my core tools, easily diagnose and fix issues and bug. It lets us stay focused, lean, and continually moving forward. The AI tools are updated regularly so I have access to the latest models and the pricing is easy to control so I can use more compute when I need it and pay extra, or reign in my spending when I need to manage my budget more closely. Review collected by and hosted on G2.com.

There are some additional integrations I'd like to see, but overall, there's nothing I dislike. Review collected by and hosted on G2.com.

The file context indexing is insanely good. I only use the auto mode, but even in that mode the way it refactors, documents, and generates boilerplate is about 85% of what I’d want. Review collected by and hosted on G2.com.

Cursor is sometimes unresponsive, which I assume is due to high traffic on their servers. It doesn’t happen often, but it’s still kind of annoying when it does.

Also, the IDE’s features sometimes just stop working mid-way. For example, the hover feature that’s supposed to show comments on a function in a formatted UI will sometimes not work at all—hovering shows nothing. Review collected by and hosted on G2.com.