What you’ll know by the end of this check

What an integrated terminal inside Claude Code desktop actually lets you do
Why this isn’t the same as running claude inside iTerm
The specific workflows that changed the day you turned this on

The shortest possible answer

The redesigned desktop app ships real integrated terminals — not a pretend terminal, a real shell — inside the Claude Code window. You can run commands, watch output, pipe results, all without leaving the app.

The shift: you stop Cmd-Tabbing between Claude Code and iTerm. That’s most of the value.

Why this isn’t just “run claude in iTerm”

The terminal that lives inside Claude Code has two things the separate terminal doesn’t:

Shared context with the agent. Commands you run and their output are visible to Claude. You type pnpm test, the tests fail, Claude already has the stack trace — no copy-paste, no re-running, no “can you see the error?”
Coexistence with agent workflow. Claude is editing a file. You notice the linter flagging something. You run the fixer in the same window, watch it land. No tab-switching, no window juggling.

Separate terminal = Claude and you are in different rooms. Integrated terminal = same room.

Workflows that change the day you turn this on

Test + iterate loops. Claude edits. You run the tests. If red, Claude already sees the output and proposes the fix. If green, you keep going. No copy-paste.

Build verification. Claude ships a change. You run the build in the integrated terminal. If it fails, Claude diagnoses without you needing to describe anything.

Scratch-pad commands. Quick gh pr view, quick jq over a response, quick curl to hit a staging endpoint. No window switch.

Shared debugging. You’re poking at something weird. Run the command. Claude reads the output directly and reasons about it.

What it doesn’t solve

It doesn’t replace your full terminal. If you have a tmux setup, seven tabs of logs, and a dedicated debugger running, that’s still its own environment. The integrated terminal is for workflow-adjacent commands — tests, builds, git, gh, quick diagnostics.

Use the integrated terminal for the commands that are part of the agent loop. Use your main terminal for everything else.

Things to try right now (6 minutes)

Open the Claude Code desktop app.
Start a session in a project repo.
Open the integrated terminal. Run git status, then ask Claude “what changed here?”
Confirm it responds with specific files (proving it saw the output).
Run a quick failing command (e.g. pnpm lint on a repo you know has one lint issue). Ask Claude to fix it.

The canonical version

Verify behavior against Anthropic’s official release notes.

Ready to verify this check?

You’ve run commands in the integrated terminal and watched Claude reference their output without copy-paste. Mark it cleared.