Critical enterprise fix: forceLoginOrgUUID and forceLoginMethod managed-settings policies were incorrectly blocking third-party provider sessions (Bedrock, Vertex, Foundry, Mantle) when used alongside an org pin — a regression introduced in v2.1.146. Enterprises on those providers should update immediately. Beyond the fix, v2.1.161 adds OTEL dimension slicing: OTEL_RESOURCE_ATTRIBUTES values are now included as labels on metric datapoints, so teams can slice usage metrics by custom dimensions like team or repo without a separate proxy. The claude agents view now shows done/total progress before the detail when work is fanned out, and the peek panel shows the longest-running item. /mcp collapses claude.ai connectors you've never signed in to behind a "Show unused connectors" row, decluttering the panel. Parallel tool calls are more resilient: a failed Bash command no longer cancels the other calls in the same batch — each tool now returns its own result independently. Linux fullscreen clipboard now uses wl-copy/xclip/xsel when available and copies to both clipboard and PRIMARY selection for middle-click paste. Several other fixes: /effort dialog and workflow animations now honor "Reduce motion"; claude -p stdout is no longer corrupted by background subagent output; /autofix-pr correctly handles git worktrees; --resume picker shows sessions from the current directory even outside a git worktree; Windows bash hooks no longer fail with "command not found."
Changelog — June 3, 2026
Three platforms shipped today: Claude Code v2.1.161 with a critical enterprise policy fix, NemoClaw's v0.0.57 documentation refresh plus several inference improvements, and an OpenAI container billing change.
The v0.0.57 release notes are now fully published in NemoClaw's docs, covering host-side sessions and agents commands, managed vLLM progress and readiness, DGX Spark model defaults, UFW auto-remediation, Slack channel validation, status failure layers, and installer tag pinning. Several inference improvements land alongside: vLLM local sandboxes now auto-detect the real context window from /v1/models.max_model_len during onboard instead of using NemoClaw's default — generated OpenClaw config now reflects the actual vLLM server limit. DeepSeek V4 Pro and Kimi K2.6 now work correctly with fetch-based OpenAI-compatible requests, fixing broken Discord and WeChat channel traffic for those models. A long-standing silent behavior is corrected: nemoclaw connect used to silently revert your model-route changes to the recorded gateway state; it now prints a loud warning naming the mismatch and the correct command to change routes, even in --probe-only mode. Installer documentation is corrected: NEMOCLAW_INSTALL_TAG must precede curl, not be placed before bash — the installer now fails with a clear error if the requested ref doesn't exist, preventing silent fallback to lkg. Managed vLLM model downloads now stream native Hugging Face progress output for visibility, and vLLM launch now polls the /v1/models readiness endpoint instead of parsing log markers.
Starting June 2, 2026, OpenAI container sessions are billed per-minute with a 5-minute minimum, replacing the previous flat 20-minute session rate. The underlying per-minute rate is unchanged — this is a billing granularity improvement. Short-lived container tasks now cost proportionally less: a 6-minute session that previously billed for 20 minutes now bills for 6. No changes to the API, SDKs, or non-container workloads.
Browse the full changelog index for the complete history across all platforms, or the daily one-liner for the most recent state of each agent.