Cursor is a consumer digital software subscription and AI-powered code editor. The platform operates as a specialized fork of Visual Studio Code, integrating large language models directly into the developer workspace. By embedding artificial intelligence into the interface, the service automates code drafting and debugging.
Cursor functions as a desktop application client that queries external AI models via proprietary APIs. The company is privately held and partners with model providers like OpenAI and Anthropic to supply intelligence. It earns revenue by charging monthly subscription fees and selling high-speed query upgrades.
How is Cursor structured as a software service?
Cursor is structured as a client-server desktop application built on the open-source VS Code editor foundation. The desktop client runs locally on the user’s computer, maintaining full compatibility with existing VS Code extensions and settings. It does not run as a web-only cloud application or a remote virtual machine interface. Instead, it acts as a local editor that forwards contextual prompts to remote servers.
The application database operates on cloud servers that index the user’s local codebase deterministically. This indexing process creates vector embeddings of local files, saving them in a local database file. When a developer queries the AI, the system matches the prompt against the codebase embeddings. The system then transmits the relevant file snippets to Cursor’s remote API clearinghouse.
The remote clearinghouse coordinates queries with third-party model providers, including OpenAI and Anthropic. These servers operate under secure data processing agreements to manage prompt traffic and verify active subscriptions. To protect developer privacy, the company does not log code snippets when privacy mode is enabled. This structure ensures that intellectual property is isolated from model training pipelines.
How do the Cursor AI editing features function?
Cursor AI editing features function by processing contextual workspace data and generating inline code edits. The platform provides codebase-wide chat, Copilot++ inline completions, and multi-file code editing modes. Codebase-wide chat allows the developer to ask questions about the entire repository index. The system automatically identifies relevant files and imports them as prompt context.
Inline completions, named Copilot++, monitor cursor movements and keystrokes in real-time. The system predicts the next edit and displays suggested changes in gray placeholder text. Users can accept these suggestions by pressing the tab key, which writes the edits directly. This predictive completion engine operates asynchronously, minimizing typing latency during active coding.
As of the mid-2026 update, the platform features Composer 2.5, an upgrade that improves sustained execution on long-running repository edits. This model employs Targeted Reinforcement Learning (RL) with textual feedback. This mechanism inserts contextual hints at the exact point in a trajectory where a model error (such as a bad tool call) occurs, correcting localized behaviors. Composer 2.5 is trained on 25 times more synthetic tasks (such as feature deletion and reimplementation) compared to version 2.0. It scores 79.8% on the SWE-Bench Multilingual and 63.2% on CursorBench v3.1 benchmarks.
What pricing tiers and referral discounts does Cursor apply?
Cursor utilizes a monthly subscription pricing model structured around usage-based credit pools and regional billing. The service is organized into four main tiers for individual developers: Hobby, Pro, Pro+, and Ultra. The Hobby tier is free of charge, providing limited agent requests and basic autocomplete tab completions. Paid individual tiers start at $20 per month for the Pro tier, expanding to $60 for Pro+ and $200 for Ultra.
| Individual Subscription Tier | Monthly Base Price | Included Monthly Credit Pool | Primary Target Use Case |
|---|---|---|---|
| Hobby | Free | Limited credits | Basic platform evaluation |
| Pro | $20 | ~$20 credit pool | Standard daily development |
| Pro+ | $60 | ~$70 credit pool | Intensive daily agent usage |
| Ultra | $200 | ~$400 credit pool | Continuous full-time agent execution |
Each paid subscription tier awards a monthly pool of usage credits denominated in dollars. These credits are consumed based on token length and the complexity of the selected AI model. Subscribers can configure pay-as-you-go overage billing to continue querying models at standard API rates after using their pool. If overages are disabled, the system defaults to a slow-pool queue with throttled processing during peak hours.
[!TIP] Cursor Discount Referral Code: New individual customers who register an account through the Cursor Referral Discount Link receive a 50% discount on their first month. This promotional coupon is valid globally for the Pro, Pro+, and Ultra subscription tiers, reducing the first month’s cost to $10, $30, or $100 respectively.
The platform processes payments recursively through credit cards, debit cards, Apple Pay, or Google Pay. Annual billing configurations are available for all paid tiers, providing a 20% discount on the cumulative monthly price. If a recurring monthly subscription payment fails, the system automatically suspends the credit pool benefits. No partial refunds are issued for cancellations made mid-month.
Who can use Cursor and what are the account limits?
Eligibility to open a Cursor account is open to developers globally with internet connectivity. The platform supports macOS, Windows, and Linux operating systems, installing as a standard local binary. Users must register an account and complete email verification to activate the editor’s AI capabilities. Commercial use is permitted across all paid tiers, complying with standard software licensing rules.
Account limits are enforced through monthly credit quotas rather than fixed daily query ceilings. The Pro tier limits fast priority requests to the initial credit pool equivalent to 500 standard queries. Priority queries receive rapid server processing, delivering completions within three seconds. Once the credit pool is exhausted, the system defaults to the slow query queue during peak demand hours.
The slow query pool is unlimited in volume but subject to queuing delays during high-traffic windows. Subscribers can buy additional credit packages to restore priority processing speeds. These add-on packs cost $20 for additional usage credits, billed to the active payment method. Daily limits do not apply to standard usage, preventing sudden mid-day transaction locks.
What are the tradeoffs and technical risks of using Cursor?
Tradeoffs of using Cursor include data privacy risks, internet dependencies, and local hardware resource demands. Because the AI features query remote servers, users must transmit code snippets across the internet. While the platform offers a privacy mode, companies with strict security policies may prohibit third-party integrations. This data transfer model represents a compliance risk for enterprise development teams.
The editor requires constant internet connectivity to process AI queries and verify subscription status. If remote server APIs experience outages, the advanced features become unavailable. The application falls back to standard offline text editing, removing the predictive completion engines. Additionally, local indexing and vector generation demand significant CPU and memory resources.
The digital support system does not operate physical offices or direct customer service phone lines. Users must submit support requests through online forums or email tickets. This digital-only troubleshooting process can take several business days to resolve account issues. Finally, subscription changes or cancellations do not refund fees paid for the active billing cycle.
Common questions
How do I cancel my Cursor subscription?
You can cancel your subscription through the billing settings page on the Cursor website. Your Pro benefits will remain active until the end of the current billing cycle. The platform does not charge cancellation fees or apply lock-in contracts.
Can I use my own OpenAI API keys?
The platform allows users to enter custom OpenAI or Anthropic API keys in the settings panel. This configuration bypasses the monthly fast query limits, billing usage directly to your personal API account. However, using personal keys may disable specific custom-trained autocomplete models.
Does the referral discount apply to existing users?
The 50% referral discount is restricted exclusively to new customers opening a paid tier for the first time. Existing accounts and returning subscribers are ineligible for the promotional first-month rate. The referral link must be clicked prior to creating the account to apply the discount.
