Claude-Code v2.1.97 Release: NO_FLICKER Improvements, Permission Fixes, and MCP Updates

Release Overview
Claude-Code v2.1.97 includes significant improvements to the NO_FLICKER mode interface, permission system hardening, and fixes for MCP (Model Context Protocol) connections. This release addresses memory leaks, transcript persistence issues, and various UI bugs across different terminal environments.
Key Features and Changes
NO_FLICKER Mode Enhancements
- Added focus view toggle (
Ctrl+O) showing prompt, one-line tool summary with edit diffstats, and final response - Fixed copying wrapped URLs inserting spaces at line breaks
- Fixed scroll rendering artifacts when running inside zellij
- Fixed crash when hovering over MCP tool results
- Fixed memory leak where API retries left stale streaming state
- Fixed slow mouse-wheel scrolling on Windows Terminal
- Fixed custom status line not displaying on terminals shorter than 24 rows
- Fixed Shift+Enter and Alt/Cmd+arrow shortcuts not working in Warp
- Fixed Korean/Japanese/Unicode text becoming garbled when copied on Windows
Permission System Fixes
- Fixed
--dangerously-skip-permissionsbeing silently downgraded to accept-edits mode after approving a write to a protected path - Hardened Bash tool permissions, tightening checks around env-var prefixes and network redirects
- Reduced false prompts on common commands
- Fixed permission rules with names matching JavaScript prototype properties (e.g.
toString) causingsettings.jsonto be silently ignored - Fixed managed-settings allow rules remaining active after admin removal until process restart
- Fixed
permissions.additionalDirectorieschanges in settings not applying mid-session - Fixed removing a directory from
settings.permissions.additionalDirectoriesrevoking access to the same directory passed via--add-dir
MCP Connection Improvements
- Fixed MCP HTTP/SSE connections accumulating ~50 MB/hr of unreleased buffers when servers reconnect
- Fixed MCP OAuth
oauth.authServerMetadataUrlnot being honored on token refresh after restart, fixing ADFS and similar IdPs - Fixed 429 retries burning all attempts in ~13 seconds when the server returns a small
Retry-After— exponential backoff now applies as a minimum
Status Line and Workspace Updates
- Added
refreshIntervalstatus line setting to re-run the status line command every N seconds - Added
workspace.git_worktreeto the status line JSON input, set when the current directory is inside a linked git worktree - Added ● N running indicator in
/agentsnext to agent types with live subagent instances
Other Notable Fixes and Improvements
- Added syntax highlighting for Cedar policy files (
.cedar,.cedarpolicy) - Fixed rate-limit upgrade options disappearing after context compaction
- Fixed several
/resumepicker issues including cache misses, lost mid-turn input, and file-edit diffs disappearing for files larger than 10KB - Fixed messages typed while Claude is working not being persisted to the transcript
- Fixed prompt-type
Stop / SubagentStophooks failing on long sessions - Fixed subagents with worktree isolation or
cwd:override leaking their working directory back to the parent session's Bash tool - Fixed compaction writing duplicate multi-MB subagent transcript files on prompt-too-long retries
- Fixed
claude plugin updatereporting "already at the latest version" for git-based marketplace plugins when the remote had newer commits - Fixed slash command picker breaking when a plugin's frontmatter
nameis a YAML boolean keyword - Fixed Bedrock SigV4 authentication failing when
AWS_BEARER_TOKEN_BEDROCKorANTHROPIC_BEDROCK_BASE_URLare set to empty strings (as GitHub Actions does for unset inputs) - Improved Accept Edits mode to auto-approve filesystem commands prefixed with safe env vars or process wrappers (e.g.
LANG=C rm foo,timeout 5 mkdir out) - Improved auto mode and bypass-permissions mode to auto-approve sandbox network access prompts
- Improved sandbox:
sandbox.network.allowMachLookupnow takes effect on macOS - Improved image handling: pasted and attached images are now compressed to the same token budget as images read via the Read tool
- Improved slash command and
@-mention completion to trigger after CJK sentence punctuation
Who This Affects
This release is relevant for developers using Claude-Code with NO_FLICKER mode, working with MCP connections, managing complex permission configurations, or experiencing issues with transcript persistence and subagent management.
📖 Read the full source: GitHub Claude-Code
👀 See Also

Claude VS Code Extension Broken on Windows After Hardcoded Linux Path in Recent Update
Anthropic's recent VS Code extension update hardcodes a Linux path, breaking the extension on Windows. Downgrading to the previous version restores functionality.

OpenClaw Hosts Its First AMA: Insights into AI Coding Agents
OpenClaw, a prominent figure in AI coding agents, hosted its first AMA on Reddit. The discussion shed light on its impacts, future plans, and challenges.

Reddit Discussion on Claude's Impact on MVP Development and Founder Pitfalls
A Reddit user discusses how Claude AI lowers technical barriers for building MVPs from $3k-$5k to DIY, but warns about increased competition and founders focusing too much on building versus marketing, PMF, and operations.

xAI loses legal challenge to California AI data disclosure law
xAI has lost its attempt to block California's AI data disclosure law, which requires companies to disclose training data sources and other details about their AI systems. The court ruling means the law will proceed as scheduled.