Anthropic's Claude Code Finally Bridges Terminal and Desktop Sessions
Claude Code's desktop app now lets you continue any CLI session with a simple /resume command, preserving the full conversation and context.
- Claude Code desktop app now supports
/resumeto continue CLI sessions with full context intact. - Feature closes the long-standing gap flagged in GitHub issue #50067.
- Sessions are stored locally as JSONL transcripts under
~/.claude/projects/. - Resume restores history, model, agent, permission mode, active goals, and scheduled tasks.
- Flags like
--mcp-config,--settings, and--add-dirmust be passed again on resume. - Desktop, web, and VS Code extension still keep separate session histories per surface.
Anthropic just closed one of the more annoying gaps in the Claude Code workflow. The desktop app can now pick up any session you started from the terminal, with the full conversation, tool calls, and working context carried over untouched.
The feature is exposed as a slash command inside the app. Type /resume, choose the CLI session you want from the picker, and you land back in the same conversation, only now inside the graphical client instead of your terminal emulator.
Closing the CLI-to-desktop gap
Until now, the CLI and the desktop app kept two parallel worlds. Users had flagged the split as a core workflow gap. The CLI supported claude --resume, claude --continue, and /resume, but the desktop app had no equivalent picker, so anyone hopping between the two lost their prior work when moving over.
Claude Code stores every session locally as a JSONL transcript at ~/.claude/projects/<project>/<session-id>.jsonl, so the raw material was always there. This update finally surfaces those transcripts inside the desktop UI.
What actually gets restored
A resumed session isn't a text replay. According to the docs, several pieces of state come back with it:
- The full conversation history, including tool calls and their results
- The model the session was running on, unless it has been retired or overridden
- The agent, keeping its system prompt, tool restrictions, and model
- The permission mode the session was in, with some exceptions covered in the docs
- Any active goal that was still running when the session ended
- Scheduled tasks that haven't expired yet
A few things do not carry over automatically. Flags like --mcp-config, --settings, --plugin-dir, --fallback-model, and directories added with --add-dir need to be passed again on resume. Standard settings files are re-read at launch, so anything living in settings.json stays applied.
Driving it from either side
Inside the desktop app, run /resume to open a session picker of your prior CLI conversations. On the CLI side, the equivalents remain the same:
| Command | Behavior |
|---|---|
claude --continue | Resumes the most recent interactive session in the current directory |
claude --resume | Opens the session picker |
claude --resume <name> | Resumes a named session directly |
/resume | Switches conversations from inside an active session |
Naming sessions with claude -n auth-refactor or /rename makes them much easier to find later, especially if you juggle several tasks in parallel.
Where this lands
The desktop app, Claude Code on the web, and the VS Code extension each still maintain their own session history, so this update is specifically about bridging CLI transcripts into the desktop client rather than unifying every surface. For anyone who does exploratory work in the terminal but prefers the desktop UI for longer reviews, that bridge is the main win. No pricing changes, no new model, no separate rollout gate. If your Claude Code desktop app is up to date, /resume is already there.
Comments
