Source Intelligence

DisclaimerUnofficial, and not affiliated with Anthropic. Nearly all of this is read straight out of what ships: npm bundles, captured prompts, published docs. Anthropic's own notes go in verbatim, marked as theirs. The rest is my reading, and every entry carries the strings behind it. If one looks wrong, vote it down and say why.

All of v1.0.48 Home All releases olderv1.0.47 v1.0.49newer
Claude Code v1.0.48

New Features

Tool Permission Modes

Added new permission modes for tool execution, giving users more control over how Claude interacts with their system:

  • default - Standard mode with confirmation prompts for potentially sensitive operations
  • plan - Plan mode where Claude outlines steps before executing
  • acceptEdits - Automatically accepts file edits without confirmation
  • bypassPermissions - Bypasses permission prompts for all tool operations

You can switch between modes during a conversation to control the level of interaction required.

PreCompact Hook

A new PreCompact hook event that runs before conversation compaction occurs. This allows custom scripts to:

  • Modify custom instructions before compaction
  • Block compaction by returning exit code 2
  • Provide feedback to users about the compaction process
Example usage
# In your hooks configuration
# Exit code 0: stdout becomes custom compact instructions
# Exit code 2: blocks compaction
# Other codes: shows stderr to user but continues

Enhanced Spinner Messages

Added support for custom spinner words/phrases that display during long-running operations, providing more engaging feedback while waiting for tasks to complete.

See this entry in the whole of v1.0.48 →