30+ fixes boost MCP stability. Loop monitoring & org-managed settings subtle but practical
Claude Code v2.1.243: Enhanced loop management and model picker customization
Original: v2.1.243
Importance: マイナーバージョンアップでバグ修正が主体。企業向けの管理設定拡張と MCP 安定性向上は実装者に有用だが、破壊的変更なし。
Summary
Claude Code v2.1.243 introduces loop execution monitoring in /usage (per-loop run count, tokens per run), organization-managed model picker customization, prompt cache TTL settings (1-hour main + 5-min subagent), organization-contracted pricing in /cost telemetry, and keyless Console sign-in. Over 30 bug fixes address MCP server auto-reconnection, GitHub connection status, cross-session messaging in containers, hook condition evaluation with command substitution, and API timeout handling (3-min timeout + retry). Enterprise flexibility and stability significantly improved.
Key Points
- Loop monitoring in /usage: run count, token consumption visibility
- Model picker and prompt cache settings now organization-managed
- MCP server auto-reconnection and reliable error reporting
- API timeout handling: ~3-min with single retry
- GitHub connection status display, improved CLI session stability
View developer notes (APIs, breaking changes, migration)
Implementation adds `promptCacheTtl` and `subagentPromptCacheTtl` settings: 1-hour main conversation + 5-min subagent cache. `modelPricing` managed setting applies org-contracted per-model rates & discount multiplier to /cost, status, telemetry instead of list price. MCP servers in non-interactive (`-p`) and SDK sessions now auto-reconnect or report failure after dropped connection. Hook conditions like `Bash(cat *)` no longer misfire when `$()` or backticks precede additional arguments. API response timeout enforced ~3min with single retry, then shows `API Error: No response from API`. Workload identity federation now pools exchanged tokens per CI job instead of re-exchanging single-use tokens.
Source: https://github.com/anthropics/claude-code/releases/tag/v2.1.243
Outlet: Claude Code Releases
This article is an AI-generated summary (OpenAI GPT-4o-mini) of publicly available information from Anthropic, OpenAI, Google, Meta, Mistral, DeepSeek, Sakana, and other vendors. The original source URL is always provided in accordance with fair-use citation requirements. Summaries are AI-generated and may contain mistranslations or misinterpretations. Always verify details with the original source.