Claude Code v2.1.193: New Shell Classification, Telemetry, and Fixes
Anthropic just shipped Claude Code v2.1.193 with a handful of quality-of-life improvements, telemetry changes, and bug fixes. Here's what you need to know.
New Settings & Telemetry
The big addition is autoMode.classifyAllShell — when enabled, every Bash/PowerShell command goes through the auto-mode classifier, not just arbitrary-code-execution patterns. This gives you finer control over shell safety.
On the telemetry side, there's a new OpenTelemetry log event: claude_code.assistant_response. It includes the model's response text, but is redacted by default. Respects OTEL_LOG_USER_PROMPTS unless you explicitly set OTEL_LOG_ASSISTANT_RESPONSES=1. To keep prompts-only, set OTEL_LOG_ASSISTANT_RESPONSES=0.
Bash Mode & MCP Improvements
Bash mode (!) now has live file path autocomplete — handy for quick commands. MCP servers that need authentication get a startup notice pointing you to /mcp. Also, the headersHelper auth helper now automatically re-runs and reconnects when a tool call returns 401 or 403.
Background Agents & Memory Management
Idle background shell commands are now automatically reaped under memory pressure. Disable with CLAUDE_CODE_DISABLE_BG_SHELL_PRESSURE_REAP=1. Background agents no longer tell Claude to "end your response" — it can keep working on other tasks while the agent runs.
Fixed: pinned background agents were being re-prompted to "Continue from where you left off" after every auto-update. Also fixed a phantom "general-purpose (resumed)" subagent spawning when backgrounding the main turn.
Other Fixes
/model and other client-data-gated UIs no longer show stale/empty state immediately after /login. Backgrounding (←←) no longer spuriously cancels with "N background tasks would be abandoned" when all tasks carry over. Agent panel now shows sibling agents when viewing a subagent. Plugin auto-rename now follows marketplace renames maps automatically. /add-dir message improved when the directory is already a working directory.
📖 Read the full source: GitHub Claude-Code
👀 See Also
Claude Cowork Now Available on Windows with Local File Access and Task Scheduling
Claude Cowork, previously exclusive to macOS, is now accessible on Windows devices. The desktop application requires a paid Claude plan, handles larger tasks with direct local file access, and allows scheduling tasks to run automatically.
NHS England retreats from open source: open letter urges reversal of SDLC-8 policy
An open letter with 74 signatures calls on NHS England to withdraw SDLC-8 — a policy that hides all NHS source code — and to reaffirm Principle 12 of the NHS Service Standard: 'Make new source code open.'
Qwen 3.6 27B Benchmarked on DeepSWE: 2% Score, 70 Hours, 44k Avg Output Tokens
Qwen 3.6 27B (FP8, BF16 KV cache, 262k context) scored 2% on DeepSWE in 70 hours. Output tokens averaged 44k per task — comparable to larger models like Qwen 3.6 Plus. Ran on 1x RTX6000 Pro Blackwell via RunPod.
Cannes Film Cost $500k to Make, $400k Was AI Compute Costs
A film screened at Cannes cost $500k to produce; $400k of that was spent on AI compute. A striking data point for AI agent developers building generative video pipelines.