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 releases Home olderv2.1.131 v2.1.133newer

Claude Code v2.1.132

15 entries read diff v2.1.131 → v2.1.132 Markdown

This release adds Linux sandbox configuration for custom bwrap and socat binaries, improves MCP startup handling with a new wait tool behind a rollout flag, and makes WebFetch failures more actionable by returning HTTP status details instead of opaque fetch errors. It also includes several terminal and background-session refinements, plus dark-launched /goal infrastructure that is present in the build but not enabled.

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

Changesopen

Custom Linux Sandbox Binary Paths#

What

Linux users can now configure exact bwrap and socat binary paths for sandbox and proxy bridge setup.

Usage
{
  "sandbox": {
    "bwrapPath": "/usr/bin/bwrap",
    "socatPath": "/usr/bin/socat"
  }
}
Details
  • bwrapPath controls the bubblewrap executable used for Linux sandboxing.
  • socatPath controls the executable used for HTTP/SOCKS bridge processes.
  • Claude Code validates that configured paths are executable and reports clearer errors if not.
  • This is useful on systems where these tools are installed outside PATH, or where a managed environment pins exact binary locations.
Evidence

Sandbox schema adds bwrapPath and socatPath settings (search for "Linux only: absolute path to the bwrap" and "Linux only: absolute path to the socat binary")

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.

MCP Wait For Servers [Gradual Rollout]#

Feature flag
tengu_ashen_kelp 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: on · compiled default in v2.1.132: 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 v2.1.132. It isn't a statement about your account. What a flag value here can and cannot tell you

What

Claude can wait briefly for MCP servers that are still connecting, then continue once their tools are available.

Usage
Claude calls WaitForMcpServers when an MCP server is still connecting.
Details
  • The new WaitForMcpServers tool waits for all pending servers, or for specific server names.
  • Results distinguish connected, failed, still pending, needs-auth, disabled, and unknown servers.
  • Missing-tool guidance now tells Claude when a requested MCP tool belongs to a server that is still connecting.
  • The tool is gated by tengu_ashen_kelp, so it may not be visible for every user yet.
Evidence

MCP wait tool and gated enablement (search for "WaitForMcpServers" and "The MCP server '")

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

WebFetch HTTP Error Results#

WebFetch now returns structured HTTP failure information when a server responds with an error status, including status text and Retry-After when present. This should help users distinguish an authentication-required page, a rate limit, and a normal fetch failure without guessing from a generic exception.

Evidence

WebFetch HTTP error result text (search for "The server returned HTTP" and "The response body was not retrieved")

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

Fullscreen Renderer Controls#

Terminal fullscreen behavior is more conservative and easier to override. Claude Code now recognizes CLAUDE_CODE_DISABLE_ALTERNATE_SCREEN as an explicit opt-out, and it auto-disables fullscreen on Windows-over-SSH ConPTY sessions unless CLAUDE_CODE_NO_FLICKER=1 is set.

Evidence

Fullscreen environment handling (search for "CLAUDE_CODE_DISABLE_ALTERNATE_SCREEN" and "fullscreen disabled: Windows over SSH")

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

Background Session Detach Key#

Background session instructions now consistently use Ctrl+Z to detach, replacing older Ctrl+B then d and Ctrl+Q guidance in redraw/startup messages.

Evidence

Background session prompts (search for "Ctrl+Z to detach")

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

MCP Connection Resilience#

Remote MCP retry behavior now logs recovery and stops retrying once failed remote servers recover. MCP stdio transports also detect excessive non-protocol stdout and disconnect before unbounded memory growth.

Evidence

MCP retry and stdout overflow handling (search for "[MCP] Retry: all remote servers recovered, stopping" and "StdoutOverflowError")

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.

Bash Safety Detection#

Read-only command validation gained additional shell-edge-case detection, especially around zsh recursive evaluation, array subscripts, assignment side effects, redirect expansion, and escaped brace handling. These checks reduce cases where a command that looks harmless can still execute dynamic shell code.

Evidence

Shell safety reasons (search for "zsh $name[expr]", "assignment alters command lookup/execution", and "Redirect target starts with =")

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

Long-Running Command Guidance#

The Bash tool description now explains run_in_background more directly and removes stale advice to read output later immediately. Claude Code also includes stronger guidance to avoid foreground sleep and use background commands or Monitor for waiting.

Evidence

Background command guidance (search for "run_in_background runs the command detached" and "Foreground sleep is blocked")

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

Policy Limit Cache Handling#

When policy limits are fetched and no restrictions apply, Claude Code now preserves an explicit cached empty result instead of treating it like a deleted cache file. This should reduce noisy policy-limit cache churn in environments with no active restrictions.

Evidence

Policy limit cache message (search for "Policy limits: No restrictions (cached empty)")

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

Service Availability Messages#

Daemon/service messages now distinguish unsupported installation from on-demand operation, making platform limitations clearer.

Evidence

Service messaging (search for "service install not available on this platform" and "service uninstall not available on linux")

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

1 entry

Bug Fixesopen

#

  • WebFetch now reports HTTP status responses as tool results instead of falling through to less-informative request failures (search for "The server returned HTTP")
  • MCP stdio servers that write too much non-JSON data to stdout are disconnected with a clear overflow error instead of allowing unbounded buffering (search for "wrote >" and "without a JSON-RPC message boundary")
  • Fullscreen mode is now disabled automatically for Windows-over-SSH ConPTY sessions where the alternate-screen renderer is known to misbehave (search for "fullscreen disabled: Windows over SSH")
  • Policy-limit fetches with no restrictions now cache the empty state instead of deleting the cache on the old 404 path (search for "Policy limits: No restrictions (cached empty)")
2 entries

In Developmentopen

#

Features with infrastructure added but not yet enabled. These are shipped "dark" and may become available in future versions.

Related

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

Goal Command [In Development]#

What

A new /goal command is being prepared to set a session-scoped stopping condition so Claude keeps working until a stated goal is met.

Status

Disabled/stubbed

Details
  • The implementation adds /goal <condition>, /goal clear, status rendering, and a session-scoped Stop hook.
  • Goal state records the condition, iteration count, and set time.
  • The command currently has isEnabled: () => !1, so users cannot invoke it in this build.
  • The existing /loops UI can see and manipulate related Stop-hook entries, but /goal itself remains unavailable.
Evidence

Disabled goal command implementation (search for "Set a goal — keep working until the condition is met" and "No goal set. Usage: /goal <condition>")

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

Verbatim
Official · Anthropic

Anthropic’s official release notes

Published verbatim by Anthropic for v2.1.132. 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 CLAUDE_CODE_SESSION_ID environment variable to the Bash tool subprocess environment, matching the session_id passed to hooks
  • Added CLAUDE_CODE_DISABLE_ALTERNATE_SCREEN=1 env var to opt out of the fullscreen alternate-screen renderer and keep the conversation in the terminal's native scrollback
  • Added a "Pasting…" footer hint while a Ctrl+V image paste is being read from the clipboard
  • Fixed external SIGINT (e.g. IDE stop button, kill -INT) not running graceful shutdown — terminal modes are now restored and the --resume hint is printed instead of an abrupt exit
  • Fixed an uncaught exception when the terminal is closed or SSH disconnects mid-session under the native build
  • Fixed --resume failing with no low surrogate in string when a tool error truncation split an emoji; pre-corrupted sessions are sanitized on load
  • Fixed --permission-mode flag being ignored when resuming a plan-mode session with -p --continue/--resume, and plan mode not being re-applied after ExitPlanMode within the same session
  • Fixed fullscreen mode showing a blank screen after laptop sleep/wake or Ctrl+Z/fg until the next keystroke or stream output
  • Fixed cursor landing mid-grapheme on Ctrl+E/A/K/U/arrow keys when an Indic conjunct or ZWJ emoji wraps across lines
  • Fixed vim operators corrupting text containing decomposed (NFD) accented characters
  • Fixed pasting text starting with / silently swallowing the input or triggering an unknown-command reply
  • Fixed pasting dumping stray escape sequences into the prompt when focus events or mouse-tracking reports interleave with the bracketed paste
  • Fixed mouse wheel scrolling being too fast in Cursor and VS Code 1.92–1.104 due to an upstream xterm.js bug
  • Fixed scroll-wheel handling in JetBrains IDE 2025.2 terminals (spurious arrow keys, wrong-direction events, runaway acceleration)
  • Fixed /usage Ctrl+S hanging when copying the stats screenshot to the clipboard on Linux/X11
  • Fixed /terminal-setup showing a contradictory error in Windows Terminal — Shift+Enter is natively supported there
  • Fixed /effort picker not reflecting the CLAUDE_CODE_EFFORT_LEVEL env var override
  • Fixed /status showing the wrong default model for some users
  • Fixed slash command autocomplete popup being capped at ~3–5 visible commands instead of scaling with terminal height
  • Fixed statusline context_window token counts reflecting cumulative session totals instead of current context usage
  • Fixed Alt+T (thinking toggle) not working on macOS terminals without "Option as Meta" enabled (iTerm2, Terminal.app defaults)
  • Fixed dead keyboard input on Windows after re-opening a background session from claude agents
  • Fixed unbounded memory growth (10GB+ RSS) when a stdio MCP server writes non-protocol data to stdout
  • Fixed MCP servers that connect but fail tools/list silently showing 0 tools — they now retry once and show "connected · tools fetch failed" in /mcp
  • Fixed unauthorized claude.ai MCP connectors showing as "failed" instead of "needs auth", and headless -p mode retrying non-transient 4xx connection failures
  • Improved visual consistency in slash command dialogs and /login, /upgrade, /extra-usage dialog spacing
  • Updated the /tui fullscreen startup banner to describe additional renderer benefits (lower memory usage, mouse support, auto-copy on select)
  • Fixed Bedrock and Vertex 400 errors when ENABLE_PROMPT_CACHING_1H is set
System prompt

No change to the system prompt since v2.1.131.

Claude Code, interactive mode