Source Intelligence
Follow Discord

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 releases Home ←olderv2.1.44 v2.1.47newer→

Claude Code v2.1.45

18 entries read diff v2.1.44 → v2.1.45 Markdown

This release updates the default model from Sonnet 4.5 to Sonnet 4.6 across the application, adds preview/markdown support for AskUserQuestion options to help users compare visual options, and introduces improved output truncation handling for long-running bash commands. The release also includes enhanced sandbox documentation, URL elicitation support for MCP tools, and new settings for customizing sandbox paths.

Find
Pick an entry · j / k steps through
14 entries

Changesopen

AskUserQuestion Preview/Markdown Support#

What

When asking users questions with multiple options, Claude Code can now display a markdown preview pane to show code snippets, ASCII mockups, or diagrams alongside each option.

Usage: When an option in AskUserQuestion includes a markdown field, the UI switches to a side-by-side layout with options on the left and a preview panel on the right.

Details
  • Previews are displayed in a monospace box when an option is focused
  • Supports multi-line text with newlines for ASCII art, code, or diagrams
  • Only supported for single-select questions (not multiSelect)
  • Users can also add free-text notes on their selection via the n key
Evidence

AskUserQuestion option schema now includes markdown field (search for "Optional preview content shown in a monospace box")

Strings lifted out of the shipped bundle, so the claim above can be checked against them.

User Notes on Question Selections#

What

Users can now add free-text notes to their selections when answering AskUserQuestion prompts.

Details
  • Press n to add notes on a selected option
  • Notes are captured in the response via annotations field
  • Useful for providing context about why a particular option was chosen
Evidence

New annotations schema (search for "Free-text notes the user added to their selection")

Strings lifted out of the shipped bundle, so the claim above can be checked against them.

URL Elicitation for MCP Tools#

What

MCP tools can now request that users open a URL (e.g., for OAuth flows) and wait for completion before continuing.

Details
  • Supports error code -32042 for URL elicitation requests
  • Shows the URL to the user with options to "Reopen URL", "Continue without waiting", or "Skip confirmation"
  • Handles completion notifications from MCP servers when the user finishes the URL flow
  • If the user cancels, the tool receives a message explaining the cancellation
Evidence

URL elicitation handler (search for "requires URL elicitation (error -32042")

Strings lifted out of the shipped bundle, so the claim above can be checked against them.

Custom Spinner Tips#

What

A new setting allows overriding the default spinner tips with custom messages.

Usage: Configure in settings with tips (array of strings) and optional excludeDefault (boolean) to hide default tips.

Evidence

New setting description (search for "Override spinner tips. tips: array of tip strings")

Strings lifted out of the shipped bundle, so the claim above can be checked against them.

Enhanced Sandbox Path Settings#

What

New settings allow configuring additional paths for sandbox read/write permissions.

Details
  • sandboxExtraAllowWritePaths: Additional paths to allow writing within the sandbox
  • sandboxExtraDenyReadPaths: Additional paths to deny reading within the sandbox
  • sandboxExtraDenyWritePaths: Additional paths to deny writing within the sandbox
  • These merge with paths from permission rules (Read/Edit allow/deny)
Evidence

New settings descriptions (search for "Additional paths to allow writing within the sandbox")

Strings lifted out of the shipped bundle, so the claim above can be checked against them.

Agent Teams Flag#

What

A new experimental --agent-teams CLI flag enables agent team functionality.

Usage
claude --agent-teams
Details
  • Sets CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 environment variable
  • Enables coordinated multi-agent workflows
Evidence

CLI flag handler (search for "--agent-teams")

Strings lifted out of the shipped bundle, so the claim above can be checked against them.

Default Model Updated to Sonnet 4.6#

The default model throughout Claude Code has been updated from Sonnet 4.5 to Sonnet 4.6, including:

  • Model selection descriptions now reference Sonnet 4.6
  • 1M context window options updated to Sonnet 4.6
  • Haiku description now compares to Sonnet 4.6 instead of 4.5
  • Plan mode hybrid option now uses "Opus 4.6 in plan mode, Sonnet 4.6 otherwise"
Evidence

Model descriptions updated (search for "Sonnet 4.6 - best for everyday tasks")

Strings lifted out of the shipped bundle, so the claim above can be checked against them.

Related

Other releases about the same thing. Found by shared names or similar wording; neither means one caused the other.

Improved Sandbox Temporary Directory Documentation#

What

The Bash tool system prompt now provides clearer guidance on using temporary directories in sandbox mode.

Details
  • Now recommends using $TMPDIR environment variable instead of hardcoded /tmp/claude
  • Notes that TMPDIR is automatically set to the correct sandbox-writable directory
  • Advises that most programs respecting TMPDIR will automatically use the correct directory
Evidence

Updated bash system prompt (search for "TMPDIR is automatically set to the correct sandbox-writable directory")

Strings lifted out of the shipped bundle, so the claim above can be checked against them.

Related

Other releases about the same thing. Found by shared names or similar wording; neither means one caused the other.

Output Truncation for Large Bash Command Output#

What

Long-running bash commands that produce large output now handle truncation more gracefully.

Details
  • Output is automatically spilled to disk when it exceeds the inline limit (8MB default)
  • Shows truncation message with total size: "Output truncated (XXX KB total)"
  • Earlier output is omitted with a message indicating how much was skipped
  • Full output is saved to a file for later retrieval
Evidence

Output truncation handler (search for "Output truncated (", "KB of earlier output omitted")

Strings lifted out of the shipped bundle, so the claim above can be checked against them.

Related

Other releases about the same thing. Found by shared names or similar wording; neither means one caused the other.

Enhanced Memory Recall Display#

What

Improved display text for memory recall operations.

Details
  • Changed from "Recalled" to "Potentially relevant memory" for proactive memory suggestions
  • Clearer distinction between recalled memories and potentially relevant memories
Evidence

New memory display strings (search for "Potentially relevant memory:")

Strings lifted out of the shipped bundle, so the claim above can be checked against them.

Related

Other releases about the same thing. Found by shared names or similar wording; neither means one caused the other.

Improved Mobile Integration Messaging#

What

Updated messaging for mobile integration feature.

Details
  • Changed from "Use /mobile to get Claude on your phone" to clearer phrasing
  • New messages: "Code everywhere with the Claude app" and "Continue coding in the Claude app"
  • QR code instructions now include keyboard shortcut: "space for QR code"
Evidence

Mobile integration strings (search for "/mobile to use Claude Code from the Claude app on your phone")

Strings lifted out of the shipped bundle, so the claim above can be checked against them.

Extra Usage Billing Labels#

What

Added clearer labels for when usage is billed as extra usage or at premium rates.

Details
  • Shows "Billed as extra usage" indicator for 1M context models when applicable
  • Shows "Billed at premium rate" for premium tier usage
Evidence

Billing label strings (search for "Billed as extra usage")

Strings lifted out of the shipped bundle, so the claim above can be checked against them.

Effort Setting Description Updated#

What

The Opus 4.6 effort setting description has been expanded to better explain the trade-offs.

Before: "Effort lets you control thinking intensity in Opus 4.6." After: "Effort lets you control the trade off between response thoroughness and token efficiency for Opus 4.6. Higher effort levels offer higher capability, and lower levels tend to use fewer tokens to optimize..."

Evidence

Updated effort description (search for "Effort lets you control the trade off between response thoroughness")

Strings lifted out of the shipped bundle, so the claim above can be checked against them.

Plugin Path Variable Handling on Windows#

What

Fixed handling of ${CLAUDE_PLUGIN_ROOT} variable replacement on Windows to correctly handle backslash path separators.

Evidence

Path replacement function now converts backslashes to forward slashes on Windows (search for "win32" ? q.replace(/\\\\/g, "/")")

Strings lifted out of the shipped bundle, so the claim above can be checked against them.

Related

Other releases about the same thing. Found by shared names or similar wording; neither means one caused the other.

1 entry

Bug Fixesopen

#

  • Fixed potential issue with semver comparison functions to use Bun's native implementation when available (search for "Bun.semver")
  • Removed dependency on js-yaml library (many YAML-related error strings removed from the codebase)
  • Fixed JSON schema validation code paths (removed $ref pointer error handling that was causing issues)
  • Improved rate limit event handling in SDK message adapter (search for "Ignoring rate_limit_event message")
1 entry

Notesopen

#

This release removes several internal dependencies including the full js-yaml library and json-schema-ref-parser, replaced with more streamlined implementations. Users upgrading from versions using Sonnet 4.5 will automatically use Sonnet 4.6 as the default model.

Related

Other releases about the same thing. Found by shared names or similar wording; neither means one caused the other.

Verbatim
Official · Anthropic

Anthropic’s official release notes

Published verbatim by Anthropic for v2.1.45. Text is unmodified from the upstream changelog. Everything else on this page came out of the bundle instead, which is why the two lists don't match.

  • Added support for Claude Sonnet 4.6
  • Added support for reading enabledPlugins and extraKnownMarketplaces from --add-dir directories
  • Added spinnerTipsOverride setting to customize spinner tips — configure tips with an array of custom tip strings, and optionally set excludeDefault: true to show only your custom tips instead of the built-in ones
  • Added SDKRateLimitInfo and SDKRateLimitEvent types to the SDK, enabling consumers to receive rate limit status updates including utilization, reset times, and overage information
  • Fixed Agent Teams teammates failing on Bedrock, Vertex, and Foundry by propagating API provider environment variables to tmux-spawned processes (anthropics/claude-code#23561)
  • Fixed sandbox "operation not permitted" errors when writing temporary files on macOS by using the correct per-user temp directory (anthropics/claude-code#21654)
  • Fixed Task tool (backgrounded agents) crashing with a ReferenceError on completion (anthropics/claude-code#22087)
  • Fixed autocomplete suggestions not being accepted on Enter when images are pasted in the input
  • Fixed skills invoked by subagents incorrectly appearing in main session context after compaction
  • Fixed excessive .claude.json.backup files accumulating on every startup
  • Fixed plugin-provided commands, agents, and hooks not being available immediately after installation without requiring a restart
  • Improved startup performance by removing eager loading of session history for stats caching
  • Improved memory usage for shell commands that produce large output — RSS no longer grows unboundedly with command output size
  • Improved collapsed read/search groups to show the current file or search pattern being processed beneath the summary line while active
  • [VSCode] Improved permission destination choice (project/user/session) to persist across sessions
Read it on GitHub →
System prompt

The system prompt was not captured for this release, so this page cannot say whether it moved.

Which releases were captured →