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 ←olderv2.1.6 v2.1.8newer→

Claude Code v2.1.7

14 entries read diff v2.1.6 → v2.1.7 Markdown

This release introduces significant speculation execution infrastructure for predictive tool running, adds a new /color command for session-customizable prompt bar colors, implements Opus 4.5 migration handling for Pro users, and improves compaction reliability with streaming retry logic. Enhanced telemetry and crash handling have also been added.

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

Changesopen

/color Command#

What

Set a custom color for the prompt bar during your current session.

Usage
/color cyan
Details
  • Available colors are configurable per session
  • Color is persisted in the session file and can be used to visually distinguish multiple concurrent Claude Code sessions
  • Sets the standaloneAgentContext.color property
Evidence

VQ9 at line 466161 (color command handler, contains "Set the prompt bar color for this session")

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

Speculative Execution Infrastructure#

What

Claude Code can now speculatively execute tool calls while waiting for user input, using isolated git clones to avoid affecting your working directory.

Details
  • Creates temporary speculation-clones/speculation-{id} directories for isolated execution
  • Uses git clone --shared --no-checkout for fast, lightweight cloning
  • Tracks time saved and tools executed via tengu_speculation telemetry
  • Changes from accepted speculations are copied back to the main worktree
  • Read operations outside git root are allowed; writes outside git root are blocked
  • Speculation aborts cleanly on user input or errors
  • Displays time saved indicator (e.g., ▶▶ 5s) when speculation completes
Evidence

g19() at line 460815 (speculation entry point, contains "speculation-clones")

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

Opus 4.5 Pro Migration Notification#

What

Pro users who hadn't set a model preference are automatically migrated to Opus 4.5, with a notification displayed.

Details
  • Only triggers for first-party auth users
  • Shows "Model updated to Opus 4.5" notification on next session start
  • Migration only occurs once, tracked via opusProMigrationComplete state
Evidence

ww9() at line 539875 (migration logic), Xq9() at line 533528 (notification display, contains "Model updated to Opus 4.5")

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

Chrome Browser Skill#

Feature flag
tengu_chrome_auto_enable On for this account, and not off by default

The flag server returned on for the one account this site reads, and nothing in this release compiles it off by default. The compiled default is shown below, and says which it is when we cannot read one: a fifth of gates compile in a string or a number rather than on or off, and most published releases have no gate table behind them at all. No client can see what the server returns for your account.

This account: on · anonymous baseline: off · compiled default in v2.1.7: no gate table built for this version

These values were read against a different version of Claude Code, so treat them as the nearest reading available instead of one taken on this release.

Read once, for one account on one subscription tier, against v2.1.7. It isn't a statement about your account. What a flag value here can and cannot tell you

What

Added a new claude-in-chrome skill that enables browser automation tools via the Skill tool.

Details
  • Provides comprehensive guidance for browser automation including GIF recording, console log debugging, and dialog handling
  • Controlled by tengu_chrome_auto_enable feature flag
  • Must invoke skill before using mcp__claude-in-chrome__* tools
  • Includes instructions for tab context management and session startup
Evidence

TI9() at line 516042 (skill registration, contains "claude-in-chrome")

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

Compaction Streaming Retry#

What

Compaction now includes retry logic when streaming responses fail.

Details
  • Controlled by tengu_compact_streaming_retry feature flag
  • Retries up to 3 times with exponential backoff
  • Improved error message: "Compaction interrupted · This may be due to network issues — please try again."
  • Logs retry attempts and final failure for diagnostics
Evidence

H97() at line 449946 (compaction handler, contains "tengu_compact_streaming_retry")

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 Crash and Signal Telemetry#

What

Better crash and shutdown diagnostics through telemetry.

Details
  • Now logs shutdown_signal events for SIGINT/SIGTERM
  • Captures uncaught_exception with error name and truncated message
  • Captures unhandled_rejection for promise rejections
  • All error messages are truncated to 2000 characters
Evidence

yJ at line 317867 (signal handlers, contains "shutdown_signal", "uncaught_exception")

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.

Keybindings File Format Update#

What

Keybindings configuration now includes JSON schema and documentation references.

Details
  • New $schema field pointing to https://code.claude.com/docs/schemas/keybindings.json
  • New $docs field linking to https://code.claude.com/docs/s/claude-code-keybindings
  • Reserved (non-rebindable) keys are now filtered from generated config
Evidence

R29() at line 473546 (keybindings generator, contains "$schema")

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

Dynamic Tool Loading Diagnostics#

What

Logs discovered tools found in message history when resuming sessions.

Details
  • Scans tool_result blocks in user messages for tool names
  • Logs count of discovered tools for debugging dynamic tool loading behavior
Evidence

_Z0() at line 279909 (tool discovery, contains "Dynamic tool loading")

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

Plugin Forced Output Style#

What

Plugins can now force a specific output style for the session.

Details
  • Plugins can set forceForPlugin: true on their output style
  • If multiple plugins force styles, a warning is logged and the first is used
  • Takes precedence over user-configured output style
Evidence

sY9() at line 508208 (output style selection, contains "forceForPlugin")

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

Orphaned Thinking Block Filtering#

What

Filters out orphaned assistant thinking blocks that don't have associated content.

Details
  • Removes assistant messages that contain only thinking/redacted_thinking blocks
  • Logs filtered messages via tengu_filtered_orphaned_thinking_message telemetry
  • Prevents display issues from incomplete responses
Evidence

Su2() at line 510058 (filtering logic, contains "tengu_filtered_orphaned_thinking")

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.

Prefetch System Context Optimization#

What

System context is now prefetched more intelligently based on trust status.

Details
  • In non-interactive sessions, prefetch runs immediately
  • In interactive sessions, only prefetches if trust has been established
  • Logs prefetch status via prefetch_system_context_* telemetry events
Evidence

tR7() at line 543352 (prefetch logic, contains "prefetch_system_context")

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

Git Conflict Detection for Snapshots#

What

Worktree snapshotting now detects in-progress git operations and provides clearer errors.

Details
  • Detects cherry-pick, rebase, merge, revert, am, and bisect operations
  • Throws descriptive error: "Cannot snapshot worktree: in-progress {operation} operation detected"
  • Also detects uninitialized repositories and prompts for initial commit
Evidence

a77() at line 458650 (conflict detection, contains "cherry-picking", "rebasing")

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.

2 entries

Bug Fixesopen

#

  • Fixed "File has been unexpectedly modified" error message to be more actionable (ZRA at line 170468)
  • Improved request size error messaging for non-interactive mode (hs8() at line 278885, contains "Request too large")

Enhanced Telemetry Beta#

Status

Feature-flagged

What

Optional enhanced telemetry collection for debugging and analytics.

Details
  • Controlled by CLAUDE_CODE_ENHANCED_TELEMETRY_BETA or ENABLE_ENHANCED_TELEMETRY_BETA env var
  • Also gated by enhanced_telemetry_beta feature flag (defaults to false)
Evidence

uI0 at line 303463 (feature definition, contains "enhanced_telemetry_beta")

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.

Verbatim
Official · Anthropic

Anthropic’s official release notes

Published verbatim by Anthropic for v2.1.7. 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 showTurnDuration setting to hide turn duration messages (e.g., "Cooked for 1m 6s")
  • Added ability to provide feedback when accepting permission prompts
  • Added inline display of agent's final response in task notifications, making it easier to see results without reading the full transcript file
  • Fixed security vulnerability where wildcard permission rules could match compound commands containing shell operators
  • Fixed false "file modified" errors on Windows when cloud sync tools, antivirus scanners, or Git touch file timestamps without changing content
  • Fixed orphaned tool_result errors when sibling tools fail during streaming execution
  • Fixed context window blocking limit being calculated using the full context window instead of the effective context window (which reserves space for max output tokens)
  • Fixed spinner briefly flashing when running local slash commands like /model or /theme
  • Fixed terminal title animation jitter by using fixed-width braille characters
  • Fixed plugins with git submodules not being fully initialized when installed
  • Fixed bash commands failing on Windows when temp directory paths contained characters like t or n that were misinterpreted as escape sequences
  • Improved typing responsiveness by reducing memory allocation overhead in terminal rendering
  • Enabled MCP tool search auto mode by default for all users. When MCP tool descriptions exceed 10% of the context window, they are automatically deferred and discovered via the MCPSearch tool instead of being loaded upfront. This reduces context usage for users with many MCP tools configured. Users can disable this by adding MCPSearch to disallowedTools in their settings.
  • Changed OAuth and API Console URLs from console.anthropic.com to platform.claude.com
  • [VSCode] Fixed claudeProcessWrapper setting passing the wrapper path instead of the Claude binary path
Read it on GitHub →
System prompt

1 of 17 tool descriptions changed.

Agent SDK, print mode

Read the prompts for this release →