Cursor Blog
AI news source. Articles are auto-selected and adapted by Hamidun News editors.
Latest publications
Cursor Releases Agent SDK: Agents as Code, Deployed With a Single Command
Cursor has opened up the Agent SDK it uses to build Bugbot, Security Reviewer and Rollouts. Agents are defined in TypeScript and Markdown and shipped with the agent-sdk deploy command.
Cursor Router: Intelligent AI Model Routing for Teams, With Savings of Up to 60%
Cursor has made Cursor Router generally available. The router picks a model for each task and, in A/B tests, saved teams up to 60% at frontier-model quality.
Cursor: AI Agent Swarm Builds 80% of SQLite in 4 Hours — at a Fraction of the Old Swarm's Cost
Cursor compared its old and new multi-agent swarms on building SQLite from scratch in Rust: better coordination delivered the same quality at a fraction of the previous cost.
xAI Publishes Grok 4.5 Model Card: Benchmarks and Safety Evaluations
xAI has released the official model card for Grok 4.5, a document with the model's benchmark results and the outcomes of its safety evaluations ahead of public use.
Cursor Launches CFO Council: 88% of Companies Deployed AI, But Only 39% See EBIT Growth
Cursor creates a council of chief financial officers to figure out why, despite global AI spending of $1.5 trillion, only 39% of companies can trace its impact on EBIT.
Cursor raises team limits and introduces Premium tier for power users
New usage plans include separate pools for Composer and third-party models, new Premium tier with 5x usage and better cost control.
Cursor and SpaceXAI released Grok 4.5 — the company’s first model beyond programming
Cursor and SpaceXAI released Grok 4.5 — a flagship MoE model that for the first time covers not only programming, but also data analytics, finance, and legal work.
Cursor Finds: 63% of Opus 4.8 Max's Benchmark Wins Are Search, Not Code
Cursor discovered that 63% of Opus 4.8 Max's successful solutions on SWE-bench Pro came not from writing code, but from finding a ready-made answer in git history or public repositories.
Cursor Launches Design Mode: Edit UI with Clicks and Drawings, Agent Fixes Code
Cursor updated Design Mode — a mode where you click, draw, or speak about interface changes, and an AI agent automatically edits the source code.
Cursor Launches Organizations to Manage Multiple Enterprise Teams
The new Organizations feature allows Enterprise customers to manage multiple Cursor teams from a single place with separate budgets, security, and access rights.
Cursor Revealed Lessons from a Year of Developing Cloud AI Agents
The company shared its experience: environment quality and execution reliability determine cloud agent performance. Cursor's system already processes 50 million actions per day.
Cursor announced the third era of software development with cloud AI agents
Cloud AI agents in Cursor now operate autonomously on virtual machines and already generate a third of all pull requests. Developers are moving to a new way of working with a team of AI assistants.
Composer 2.5 from Cursor: code agent trained to handle long-term tasks better
Cursor released Composer 2.5 update with a new training method. The agent now handles long-term tasks better and follows developer instructions more accurately.
Cursor 3: A Unified Development Workspace for AI-Powered Agents
Cursor 3 is a new development interface for working with AI agents. The workspace supports parallel work of multiple agents, synchronization between cloud and local environments, an integrated browser, and fast PR creati
Cursor released Composer 2 — a frontier coding model at $0.50 per million tokens
Cursor introduced Composer 2, a new coding model that improved results across all benchmarks (CursorBench 61.3, SWE-bench 73.7) and costs $0.50 per million input tokens, with a speed-focused variant priced below competit
Cursor launched environments for cloud agents: multi-repo and versioning
Cursor released tools for configuring dev environments for cloud agents, allowing them to work with multiple repositories at once and with the full project context.
Composer autoinstall: how older versions help train new ones
Cursor uses its own models: Composer 1.5 prepares environments so Composer 2 can train more efficiently. This saves compute resources and improves the quality of the training signal.
How Cursor Makes Its AI Agent Better: From Guardrails to Dynamic Context
Cursor has revealed the methodology behind improving its coding agent: how it moved away from rigid guardrails in favor of dynamic context and measures quality through Keep Rate — the share of code that remains in the pr