Claude Code v2.1.176: Language-Aware Sessions, Bedrock Credential Caching, and Dozens of Fixes

Claude Code v2.1.176 is out, bringing language-aware session titles, better Bedrock credential caching, and a large batch of fixes — especially for Remote Control, model enforcement, and tmux clipboard behavior.
New Features
- Language-aware session titles: Session titles are now generated in the language of the conversation. You can pin a specific language via the
languagesetting. - Footer link badges via regex: New
footerLinksRegexessetting lets you add regex-matched link badges to the footer row. Configurable via user or managed settings. - Improved Bedrock credential caching: Credentials from
awsCredentialExportare now cached until theirExpirationtime, instead of a fixed 1-hour window. This reduces unnecessary re-authentication.
Model and Enforcement Fixes
- Fixed
availableModelsenforcement: Alias model picks could previously be redirected to a blocked model viaANTHROPIC_DEFAULT_*_MODELenvironment variables. This is now enforced correctly. /fastnow refuses to toggle if it would switch to a model outside the allowlist.- Auto mode fallback for Fable 5: Organizations without Opus 4.8 enabled no longer break auto mode. The classifier now falls back to the best available Opus model instead of failing.
Hooks and Sandbox
- Fix hook
ifconditions for Read/Edit/Write tool paths: Patterns likeEdit(src/**),Read(~/.ssh/**), andRead(.env)now match correctly. - Linux sandbox: Fixed a crash when
.claude/settings.jsonis a symlink with an absolute target.
Clipboard and Terminal
/copyand mouse-selection copy now work inside tmux over SSH — populating the system clipboard. Previously they failed to reach the clipboard.- tmux paste buffer now loads correctly on versions older than 3.2.
Remote Control
- Fixed silent model switching when connecting from web/mobile.
- Disconnect notifications now show a human-readable reason instead of a bare numeric code.
- Connection failures no longer add a duplicate line to the conversation transcript.
- Fixed sessions not disconnecting when signing into a different account.
Git and Worktree
/cdand worktree moves no longer leave the session reporting the previous directory's git branch.
Background Sessions and Agents
- Pressing back in one agent window no longer detaches other windows attached to the same session.
- Fixed backgrounded sessions showing "Working" forever when
/bgmid-turn had nothing left to continue. - Background agent search by PR URL now works — PRs opened during scheduled wakeups or while a job was blocked appear in
claude agentssearch. - Fixed the agents view input showing no text cursor on Windows.
claude --bg -cn <name>now correctly seeds the session name.- Windows network paths in persisted state are neutralized before respawn.
- Malformed resume IDs from corrupted state files are now rejected instead of causing crashes.
- Fixed the Windows background-service daemon not starting when
~/.claude/daemonhas the ReadOnly attribute set.
Cloud Sessions and Daemon
- Fixed cloud sessions failing with "Could not resolve authentication method" when idle for too long before being claimed.
- Background sessions now show clearer guidance when a window left open across an auto-update can't submit a reply, and
claude daemon statusexplains version-skew behavior.
As always, update via the launcher or download from the releases page.
📖 Read the full source: GitHub Claude-Code
👀 See Also

Why AI Coding Agents Churn Out Slop After 20 Turns: Context Blindness
A deep audit of API logs reveals Cursor and Claude Code aren't getting dumber — they suffocate on bloated context windows filled with noise, causing architectural destruction.

RunAnywhere RCLI: On-Device Voice AI Pipeline for Apple Silicon
RunAnywhere has released RCLI, an open-source voice AI pipeline for macOS that runs STT, LLM, and TTS entirely on Apple Silicon devices. The tool uses their proprietary MetalRT inference engine and claims significant performance improvements over existing solutions.

Markdown as Protocol for Agentic UI with Streaming Execution
A prototype uses Markdown as a unified protocol for AI agents to stream text, executable code, and data in a single response. It features streaming execution where code runs statement-by-statement as it arrives and a mount() primitive for creating React UIs with data flow between client, server, and LLM.

No-Code Persistent Memory System for Claude Using Notion and MCP
A radiologist built a 'Cognitive Hub' in Notion that Claude reads and writes to through MCP, creating a structured knowledge base with a routing table to load only relevant information per conversation. The system has grown to 70+ pages after a month of daily use.