Use Claude in your workspace
Two surfaces, one machine: the Claude Code CLI in the terminal, and Claude inside the VS Code editor.
In the terminal#
Open the Terminal tab on the machine. Claude is on $PATH:
user@dev-1:~/work$ claude> Welcome to Claude Code. What are we building?user@dev-1:~/work$ claude "explain the auth flow in this repo"Reading auth/middleware.py, auth/jwt.py, settings.py ...# ... full thinking-then-answer output ...user@dev-1:~/work$ claude --resumeResuming previous session.
In VS Code#
Open the VS Code tab. Claude is wired into the side panel - same auth, same machine, same files. The Claude pane shows the conversation; the editor shows your repo.
14 def verify_token(token): 15 try: 16 claims = jwt.decode(token, ...) 17 except jwt.ExpiredSignatureError: 18 Claude is editing here
jwt.decode call and raise TokenExpired with the expiry timestamp. Patch is ready to review.Letting Claude run commands#
Claude Code runs commands inside your machine, not on Anthropic's servers. By default it will ask permission before any write operation (running a script, creating a file, hitting the network). You can:
- Approve case-by-case: the safe default. Claude pauses on each new command type.
- Approve for the session: accept once for "always allow this session" - lets Claude rip through obvious-safe commands without nagging you.
- Approve permanently: edit
~/.claude/settings.jsonto whitelist commands you consider safe. EasyEnv ships an opinionated default for the recipe you picked.
Tokens and billing#
Token usage and the per-message billing is between you and Anthropic - EasyEnv doesn't add a markup or proxy your traffic. The machine talks directly to Anthropic over the public internet using the credentials you signed in with. Run claude /cost in the terminal to see your current session's spend.
