Maximizing IDE productivity
Cursor is a VS Code fork and integrates Large Language Models directly into the editor. You can choose between Ask Mode to ask questions or Agent Mode to allow the AI to edit code and use the terminal. While Agent Mode handles multi-file tasks, it often fixes more than you request. Use @filename or @foldername to narrow the context for better results. For specific code edits, use the CTRL + I shortcut to copy selections into the chat window. You can also write project rules in configuration files like .cursorrules to define how the AI behaves. To verify the AI processes these, you can include the instruction to output "I read the Roles" in your Project Rules. For regular tasks, many users choose cursor small because it provides unlimited use for easy tasks. For more difficult tasks, users often select Claude-3.7-sonnet. Some users prefer Claude-4-sonnet, which was released on May 23. Tab completion suggests the next block of code, and you can accept it with a single press of the Tab key. This tool can also generate entire functions from natural language instructions. The system uses Composer 2 to inspect all your code so that recommendations fit the overall context. This version can assemble logic, structure, and even edge cases for entire features.
Current subscription tiers
Pricing changed on October 12, 2026, when the company introduced The X Subscription. The tiers provide different access levels to models such as Grok, Claude, and Gemini. The Plus plan costs $8 per month and only provides access to Grok models. The Premium tier costs $30 per month and includes Grok plus access to third-party models at their respective API list prices, which could lead to higher costs if you use these models heavily throughout the month. If you require much higher usage, the Super plan costs $100 and the Ultra plan costs $200.
| Plan | Monthly Price | Model Access |
|---|---|---|
| Free | $0 | Limited Agent and Grok |
| Plus | $8 | Grok only |
| Premium | $30 | Grok plus third-party models via API |
| Super | $100 | 15x Plus usage |
| Ultra | $200 | 50x Plus usage |
The Premium plan allows for 4x the usage of the Plus tier. The Super plan provides 15x the usage of the Plus tier, while the Ultra plan provides 50x the usage of the Plus tier. You should check your specific usage needs before you commit to a plan, as third-party models are billed separately by usage, assuming you already understand the API cost implications. Existing subscribers to the old $20 Pro plan can keep their original rate. The student free offer for Pro ended on June 25, 2026. On July 28, 2026, the company launched Cursor Start in India for ₹649 per month.
Security vulnerabilities and reliability
Security teams must monitor Cursor because the tool introduces new attack vectors. In October 2025, five high-severity vulnerabilities were disclosed in versions 1.7 and below. These vulnerabilities allow remote code execution through prompt injection. One vulnerability, CVE-2025-59944, allows attackers to modify sensitive files through case-sensitivity bypasses. Another issue, CVE-2025-61590, involves VS Code workspace manipulation. A third flaw, CVE-2025-61591, involves malicious MCP server impersonation through OAuth, which can lead to full user privilege compromise on the host system.
Cursor provides significant speed for solo developers, but it requires constant human verification to prevent security flaws and logic errors. The AI tends to tell lies with high confidence, so you must confirm all outputs manually. The AI works by analyzing your entire codebase to provide context-aware suggestions. This context-awareness reduces context switches by 40% for many engineering teams. However, the AI might also produce plausible but incorrect code for niche APIs. Developers also face risks from trusted commands, such as the AI suggesting shell commands or dependency installations. Does the AI ever fully understand the project structure without human intervention? You must always verify AI-generated code against your own requirements.
