Two Claude Code releases in a matter of days — and both bring features people have been waiting for.
Version 2.1.74: Context Suggestions
The highlight of 2.1.74 is Actionable Context Suggestions. Claude Code now actively suggests what context you should add for better results. Instead of guessing which files or info Claude needs, you get concrete recommendations.
There’s also autoMemoryDirectory — Claude Code now automatically stores project context in a configurable directory. Plus fixes for a memory leak and an RTL text rendering issue.
Version 2.1.75: 1 Million Token Context
Just days later, 2.1.75 drops with the biggest upgrade: a 1-million-token context window for Opus 4.6 — enabled by default. Previously you had to configure this separately. Now it’s just there.
What this means in practice: you can load entire codebases into context without constantly hitting limits. For larger projects, this is a genuine game changer.
Also new: the /color command for customizing your Claude Code terminal color scheme. And memory entries now get timestamps — helpful when you want to track when Claude learned what.
My take
The context suggestions are smart. One of the biggest problems with AI coding tools is that users don’t know what context to provide. If Claude Code can detect and suggest that itself, it significantly lowers the barrier to entry.
And the 1M token window as default? That tells you where things are heading. Context is king — and more of it means better results.
Sources: Claude Code Changelog, GitHub Releases