Claude Code speeds up startup and opens the books on token spend
- Source
- x.com
- Date
We've been shipping a lot of improvements to Claude Code based on your feedback. This week we made CLI startup faster, added more visibility into where your tokens go, made auto mode rules easier to see and edit, and shipped even more Remote Control fixes.
The CLI now starts up faster. It no longer waits for the sandbox and MCP servers before you can start typing, and running `claude` skips setup it doesn't need. The Linux x64 download is 4.5x smaller, down to ~75 MB, and native builds use 40–70 MB less memory per session.
You asked for more visibility into where your tokens go. /cost now has a prompt-cache line, /usage breaks down your /loop tasks by runs and tokens so you can see which loops burn many tokens, and /tasks shows which model and effort level each subagent ran on.
Many auto mode questions were about what it actually allows and why it sometimes blocks a command you actually wanted. There's now an "Auto mode" tab in /permissions where you can read the defaults and add your own rules, without having to edit ~/.claude/settings.json directly.
Context
Anthropic's Claude Code update makes recurring work easier to inspect through its usage reports. The release thread says /usage now reports both execution counts and token consumption for /loop tasks. Seeing those together helps a developer distinguish frequent repeated work from an individually expensive run.
The update also exposes the model and effort level used by each through /tasks. These details connect consumption to the work and settings that produced it, giving developers more evidence when deciding which recurring tasks to adjust.
- AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
- multi-agent — Using several AI agents on one problem — splitting work in parallel, checking each other, or filling different roles like planner and reviewer.
- sandbox — An isolated environment where AI-generated code or agent actions run without being able to touch anything real.
- MCP — The Model Context Protocol — an open standard that lets any AI assistant plug into any tool or data source without custom integration code.
Checking sign-in…
Loading comments…