Cursor Updates and Changelog
How Cursor Became the Default
The Cursor changelog is relentless. Between November 2025 and May 2026, Anysphere shipped Background Agents, an in-editor code reviewer, instant grep, Jupyter notebook support, MCP server integration, a JetBrains expansion, a proprietary model (Composer 2, then 2.5), Jira integration, shareable canvases, and an auto-review mode that lets agents work longer without manual approvals. Any one of those would have been a headline feature for a slower-moving company. Cursor shipped them all in seven months.
That velocity explains the numbers. Cursor crossed $500 million in annual recurring revenue and a $10 billion valuation in 2025. Over half of Fortune 500 companies adopted it. And in April 2026, xAI struck a deal that gives SpaceX the right to acquire Anysphere for $60 billion, a figure that would have seemed absurd eighteen months earlier.
The strategic picture is clear: Cursor is no longer competing with GitHub Copilot on autocomplete. It is building toward what it calls self-driving codebases, where agents merge pull requests, manage rollouts, and monitor production autonomously. Whether that vision is two years away or ten, the product is already the closest thing to it that exists. The timeline below documents every release so you can track the actual shipping cadence against the ambition.
Auto-review is a new run mode that lets agents work longer with fewer manual approval prompts. Allowlisted calls run immediately, sandboxable calls run in the sandbox, and everything else goes to a classifier subagent that decides whether to allow, reroute, or ask for approval. The mode applies to shell, MCP, and fetch tool calls. Users can configure it in settings and provide custom instructions to the classifier. This is the clearest signal yet of Cursor's direction: make the agent more autonomous so developers spend less time babysitting it and more time reviewing its output.
Canvases are interactive artifacts that agents create during conversations: reports, dashboards, custom interfaces. With this update, developers can share a link to a live canvas snapshot that teammates can open in a browser without sharing the full chat thread. This moves Cursor's output beyond code and into the territory of collaborative documents.
Cursor is now available directly in Jira. Teams can assign work items to Cursor or mention @Cursor in a comment to kick off a cloud agent. The agent uses the ticket title, description, comments, and repository settings to scope the task. When finished, Jira shows completion updates with a link to the pull request. This completes the loop between issue tracking and autonomous code generation, following the earlier Linear integration.
A substantial intelligence upgrade over Composer 2, particularly on long-horizon agentic tasks. Composer 2.5 follows complex instructions more reliably, maintains better context over extended sessions, and delivers frontier-level coding performance at a fraction of the cost of third-party models. Cursor positioned it as the model that makes their new pricing work: frontier performance without frontier API bills. For teams evaluating whether to use Cursor's first-party model or pay extra for GPT or Claude, this release was designed to make the first-party option compelling enough that most developers stop reaching for the third-party toggle.
xAI announced it struck a deal with Anysphere, giving SpaceX the right to acquire the company for $60 billion or pay $10 billion for collaborative work. The deal valued Cursor's parent company at roughly 120x its annual revenue, reflecting the strategic importance of AI-native development tools. If exercised, the acquisition would place Cursor alongside Grok and the X platform under the SpaceXAI umbrella.
Cursor agents gained the ability to respond by creating interactive canvases: rich visual outputs like reports, dashboards, and custom interfaces generated directly in the chat panel. This extended Cursor beyond pure code editing into broader knowledge work.
Cursor launched its first proprietary coding model, Composer 2, built on Kimi K2.5. It delivered major benchmark gains in code generation tasks and represented Cursor's shift from being purely a wrapper around third-party models to having its own model stack. This reduced dependency on OpenAI and Anthropic while giving Cursor pricing leverage.
Cursor became available in JetBrains IDEs including IntelliJ, PyCharm, and WebStorm, offering AI model choice and secure code search. This opened Cursor to the large population of Java, Kotlin, and Python developers who prefer JetBrains tooling over VS Code.
A dense release. Plan mode gained clarifying questions via interactive UI. AI code review appeared in the editor itself, complementing BugBot's PR-level reviews on GitHub, GitLab, and GitHub Enterprise Server. And all agent grep commands became instant, a seemingly small improvement that dramatically sped up codebase exploration.
Memories let Cursor remember facts from conversations and apply them in future sessions, creating a persistent project knowledge base. MCP server support with one-click setup and OAuth made it possible to connect Cursor to external tools and data sources without custom integration work. Together these features transformed Cursor from a stateless assistant into a context-aware development partner.
The landmark release. Background Agents execute coding tasks independently while developers focus on other work. They can create branches, make changes across multiple files, run tests, and prepare pull requests without continuous oversight. This was Cursor's defining feature: the shift from AI as pair programmer to AI as parallel worker. Within months, over 50% of Fortune 500 companies had adopted Cursor, and annual recurring revenue crossed $500 million.
BugBot launched as an automated pull request reviewer that catches issues before merge. It integrates with GitHub, GitHub Enterprise Server, and GitLab. Flagged issues include a Fix in Cursor prompt that takes developers directly to the problematic code.
Common Questions About Cursor Updates and Changelog
How much does Cursor cost?
Cursor offers a free tier with limited completions. The Pro plan costs $20 per month for individual developers. Teams pricing is $40 per month per seat for the Standard tier, with a Premium seat option offering 5x usage at roughly 3x the cost. Pricing is shifting to separate pools for first-party Cursor models and third-party API usage.
Is Cursor better than GitHub Copilot?
Cursor has moved faster on agentic features like Background Agents, Jira integration, and autonomous task completion. Copilot has broader platform integration across the GitHub ecosystem and is available in more IDEs. Cursor generally leads on autonomous coding; Copilot leads on ecosystem breadth.
Is xAI acquiring Cursor?
In April 2026, xAI announced it secured the right to acquire Anysphere (Cursor’s parent company) for $60 billion or pay $10 billion for collaborative work. As of this writing, the acquisition option has not been exercised. Cursor continues to operate independently.