Claude Code versions 2.0.64-2.0.67 introduce several quality-of-life improvements: instant auto-compacting eliminates annoying waits, the new /stats command shows token usage, session renaming makes conversation management easier, and new connectors (MCP integrations) expand what Claude can interact with. The updated plan mode now transfers context between conversations automatically.
/rename or claude -r to make finding previous conversations much easierOption+P to change models mid-prompt without losing your workThe auto-compact feature was one of the most frustrating aspects of Claude Code for regular users. When context grew too large, compacting would interrupt your flow. In version 2.0.67, auto-compacting happens instantly in the background, making long coding sessions much smoother.
The new /stats command provides visibility into your Claude Code usage. You can see total tokens consumed, break down usage by model (Sonnet 4.5 vs Opus), and even see fun comparisons like "21 times more tokens than Lord of the Rings." The presenter's team had used 13 million tokens on Sonnet alone, demonstrating just how much professional users rely on the tool.
Finding old conversations has always been challenging when using claude -r to resume sessions. The new rename feature lets you give meaningful names to sessions like "working on grove" instead of cryptic auto-generated titles. Combined with session search, this makes navigating your conversation history much more manageable.
"Once you find one and you're in the right one, you could rename it and then you wouldn't lose that conversation. It's kind of like a save file concept - video games."
A common frustration was writing a detailed prompt only to realize you're on the wrong model. Now pressing Option+P lets you switch between Sonnet, Opus, or other models without losing your prompt text. While you could always copy-paste before, this UX improvement removes friction from the workflow.
Opus 4.5 now has thinking mode (extended thinking) enabled by default. This change addresses complaints about Opus quality - many users may have been running without thinking mode, which significantly reduces output quality for complex tasks.
"This might actually be the reason why people were talking about the quality being so bad on Opus. They weren't using thinking mode."
Anthropic added numerous connectors - essentially MCP (Model Context Protocol) integrations that let Claude interact with external services. Notable additions include Cloudflare, Canva, Figma (for design-to-code workflows), Monday, Netlify, and PayPal. These enable workflows like listing workflows, searching for actions, and executing operations directly from Claude.
The improved plan mode (Shift+Tab to toggle) creates comprehensive plans that persist across conversation compacts. This means you can let Claude create an implementation plan and watch it execute without worrying about losing context. The plan transfers between conversations automatically.
"You don't actually have to worry about compacts anymore because the plan is transferred between conversations."
"Claude Code, in my opinion, is the best coding tool on the market. It's better than Cursor. It's better than pretty much every other thing that I've ever used."
"I've used 21 times more tokens than Lord of the Rings, which is absurd."
"Thinking mode makes Claude much better, especially Opus 4.5."
| Command | Description |
|---|---|
/stats | View token usage statistics by model |
/rename [name] | Rename current session |
claude update | Update to latest Claude Code version |
claude -r | Resume a previous session |
Option+P | Switch models while writing a prompt |
Shift+Tab | Toggle plan mode |
Tab | Accept prompt suggestion |