> This page is for version Latest (default).
> For other versions, use one of these documentation indexes:
> - Latest (default): https://docs.nvidia.com/nemoclaw/latest/llms.txt

> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.nvidia.com/nemoclaw/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/nemoclaw/_mcp/server.

## v0.0.138

NemoClaw v0.0.138 tightens installation and image-build security and fixes onboarding, local inference, and timeout recovery.
The inference guides also clarify when hosted custom endpoints use native providers or the shared route.

* Linux system-mode Ollama installation verifies a pinned official installer before running it and stops if the downloaded content cannot be verified.
  The repository `.dockerignore` now excludes credential, key, and certificate names at every directory depth, including nested files in Docker build contexts.
  Related change: [PR #13614](https://github.com/NVIDIA/NemoClaw/pull/13614).
* New hosted custom endpoint selections outside the Personal tier can use sandbox-attached providers when supported.
  Fresh Personal-tier hosted selections stay on `inference.local` without a native attachment, and retained shared selections remain on their recorded route.
  The provider guides now describe these paths for OpenClaw, Hermes, Deep Agents Code, and Pi.
  Related changes: [PR #13528](https://github.com/NVIDIA/NemoClaw/pull/13528), [PR #13609](https://github.com/NVIDIA/NemoClaw/pull/13609).
* New local-inference profiles for `host.openshell.internal` and `host.docker.internal` admit IPv6 unique-local addresses as well as private IPv4 addresses.
  Existing sandboxes keep their recorded permissions until the local provider is reselected.
  Public, link-local, and loopback IPv6 addresses remain excluded.
  Related change: [PR #13582](https://github.com/NVIDIA/NemoClaw/pull/13582).
* `NEMOCLAW_STATUS_PROBE_TIMEOUT_MS` and `NEMOCLAW_SANDBOX_EXEC_TIMEOUT_MS` values above `2147483647` milliseconds now fall back to their defaults instead of timing out immediately.
  Related change: [PR #13550](https://github.com/NVIDIA/NemoClaw/pull/13550).
* Onboarding records the requested compatible-endpoint reasoning effort with the new sandbox so resume and rebuild use that selection.
  Related change: [PR #12814](https://github.com/NVIDIA/NemoClaw/pull/12814).
* Interactive onboarding keeps `Ctrl-C` working when prompt output is captured.
  Closing stdin during a prompt or between questions cancels with a nonzero exit and preserves the session for `nemoclaw onboard --resume`.
  Related changes: [PR #12210](https://github.com/NVIDIA/NemoClaw/pull/12210), [PR #12216](https://github.com/NVIDIA/NemoClaw/pull/12216).
* The Hermes managed Dockerfile pins a base image that satisfies the current security inventory.
  Related change: [PR #13581](https://github.com/NVIDIA/NemoClaw/pull/13581).