Rules generation, improved agent terminal and MCP images

Automated and improved rules

You can now generate rules directly from a conversation using the /Generate Cursor Rules command. This is useful when you want to capture the existing context of a conversation to reuse later.

For Auto Attached rules with path patterns defined, Agent will now automatically apply the right rules when reading or writing files

We’ve also fixed a long-standing issue where Always attached rules now persist across longer conversations. Agent can now also edit rules reliably.

More accessible history

Chat history has moved into the command palette. You can access it from the "Show history button" in Chat as well as through the Show Chat History command

Making reviews easier

Reviewing agent generated code is now easier with a built-in diff view at the end of each conversation. You'll find the Review changes button at the bottom of chat after a message from the agent.

Images in MCP

You can now pass images as part of the context in MCP servers. This helps when screenshots, UI mocks, or diagrams add essential context to a question or prompt.

Improved agent terminal control

We've added more control for you over terminals started by the agent. Commands can now be edited before they run, or skipped entirely. We've also renamed "Pop-out" to "Move to background" to better reflect what it does.

Global ignore files

You can now define global ignore patterns that apply across all projects via your user-level settings. This keeps noisy or sensitive files like build outputs or secrets out of prompts, without needing per-project configuration.

New models

We've recently added many more models you can use. Try out Gemini 2.5 Pro, Gemini 2.5 Flash, Grok 3, Grok 3 Mini, GPT-4.1, o3 and o4-mini from model settings.

Project structure in context (Beta)

We’re introducing an option to include project structure in context, which adds your directory structure to the prompt. The agent now gets a clearer sense of how your project is organized, which improves suggestions and navigation of large or nested monorepos.

---

Keybindings (2)

• Some CMD+K shortcuts are now remappable
• Emacs keybinding extensions now work reliably

Improvements (5)

• Simplified model picker UI for Auto-select
• New interface for command palette
• Refreshed UI for Tab jump suggestions. Suggestions outside viewport - are now clickable
• Tooltips for modes in Chat
• MCP stability
• You can now connect to AWS Bedrock using access keys and secret keys
• Git > @PR has now been renamed to @Branch

Fixes (4)

• Always-attached rules now persist correctly across summarization
• Rules in .mdc files can now be created and edited without issue by agent
• Fixed selection issues with @mention nodes at the start of the input box
• Performance in core editor

Business / Teams (4)

Global ignore traversal

• Cursor can now traverse up directory trees to find ignore files. This behavior is off by default and can be enabled from admin settings. IAM roles for Bedrock

• Enterprises can now connect to AWS Bedrock using IAM roles User-level usage insights

• Admins can now view per-user spend and fast request usage directly from the dashboard. Auto-run controls for teams

• Admins can configure global MCP settings from dashboard.

Patches (2)

0.49.1

• Fixed the Cursor Rules editor for Remote SSH
• UI delimiter fix for zsh terminal in Chat
• Rules no longer show "max length" warning when under size limit
• Fixed sporadic "message too large" warning in empty chats

0.49.2

• Improved rule generation apply
• Improved caching for user prompts
• Improved client metric handling