Release Notes

July 17, 2026

v0.0.87

NemoClaw v0.0.87 adds bounded DGX Station factory-image qualification paths, fixes Station post-reboot resume, makes managed Deep Agents Code startup restart-safe, and improves rebuild recovery, managed vLLM storage checks, sandbox backups, and strict-provider compatibility.

  • DGX Station Express recognizes the exact April 2026 NVIDIA Colossus BaseOS and June 2026 NVIDIA AI Developer Tools GB300 factory profiles for qualification. The installer preserves each factory kernel, driver, Docker, and NVIDIA Container Toolkit stack, applies only the bounded access or runtime preparation required by that exact profile, and rejects identity, package, service, GPU, or runtime drift. DGX Station remains Deferred while physical qualification continues. For more information, refer to Prepare DGX Station to Install NemoClaw and Platform Support and Launch Claims.
  • Station Express onboarding now accepts both the current six-field installer resume receipt and the legacy three-field format after host preparation requires a reboot. The current format validates the agent, sandbox name, and policy tier in addition to the pinned revision, model, and receipt generation, while malformed, unsupported, or extended receipts remain fail-closed for troubleshooting. For more information, refer to Prepare DGX Station to Install NemoClaw.
  • Managed Deep Agents Code onboarding now persists the nemoclaw-dcode-entrypoint startup command when the OpenShell Docker driver recreates a sandbox. The recreated container also receives the required nproc=512:512 and nofile=65536:65536 limits, so the managed runtime remains available after a gateway restart without weakening its process and file-descriptor boundaries. For more information, refer to Security Best Practices.
  • Rebuild recovery verifies that a restored Hermes sandbox returns to healthy gateway and managed MCP state before reporting success. OpenClaw rebuilds also clear stale managed-provider session pins after an inference switch, allowing restored sessions to use the current configured model while preserving intentional pins to other providers. For more information, refer to Recover and Rebuild Sandboxes and Switch Inference Providers.
  • Managed vLLM storage preflight estimates cold image and model downloads from pinned image metadata and model payload sizes. It checks Docker storage and the Hugging Face cache separately when they use different filesystems, rechecks capacity after a cold image pull, warns and continues during express or other non-interactive setup, and requires confirmation during interactive setup. For more information, refer to Set Up vLLM.
  • Sandbox backup creation now streams archive data and validates entries incrementally instead of buffering the complete archive in host memory. Large backups therefore retain the existing traversal checks and partial-state behavior without requiring memory proportional to the archive size. For more information, refer to Create and Restore Snapshots.
  • Hermes registers NemoClaw tools with the single function-schema envelope required by strict OpenAI-compatible providers. Google Gemini no longer rejects the managed Hermes tool list because of a nested schema, and audio transcription retains its declared parameters. For more information, refer to Use Google Gemini.
  • Replacement-image rebuild failures preserve bounded, redacted Docker diagnostics when process output arrives as buffered data, making host-specific build failures actionable without exposing credentials or private host paths. For more information, refer to Recover and Rebuild Sandboxes.

v0.0.86

NemoClaw v0.0.86 enables the Station express recipe on qualified stock DGX OS GB300 systems, makes interrupted Station setup resumable, and fixes model validation, managed vLLM cache checks, sandbox builds, and upgrade guidance.

  • DGX Station GB300 express setup now accepts stock DGX OS 7.2.0, 7.4.0, and 7.5.0 when strict Station, GB300, release-marker, driver, ECC, Docker, CDI, and GPU-container checks pass. Direct-GPU sandboxes receive only the exact read-only GPU, CPU, memory, NUMA, and NVIDIA module-initialization sysfs paths required for CUDA instead of broad /sys access. A clean physical DGX OS 7.5.0 validation completed with local Nemotron Ultra inference, sandbox CUDA, and Hermes file-tool use; DGX Station support remains Deferred pending broader qualification. For more information, refer to Prerequisites, the NemoClaw Quickstart, and Platform Support and Launch Claims.
  • Interrupted Station Express setup now persists its validated, secret-free provider, model, sandbox, and interaction choices. nemoclaw onboard --resume restores those choices and retries the failed managed-vLLM step, while successful onboarding and --fresh retire stale Express intent and installer reboot receipts. For more information, refer to the NemoClaw CLI Commands Reference and Set Up vLLM.
  • Managed vLLM cache preflight now checks only the Hugging Face cache paths used by the selected model. Root-owned artifacts from an unrelated model no longer block a model switch, and repair guidance identifies the exact unwritable path. For more information, refer to Set Up vLLM.
  • Manual Google Gemini model IDs are validated against Google’s native model catalog before the existing OpenAI-compatible chat-completions probe. Catalog results with or without the models/ prefix are normalized, non-chat models are filtered out, and API credentials remain outside process arguments. For more information, refer to Use Google Gemini.
  • Sandbox image staging normalizes script and directory permissions before Docker consumes the build context. Fresh installs created under a restrictive umask no longer carry root-only modes into later non-root image-build stages.
  • Legacy sandbox recreation now warns before the destructive step that managed recovery restores .openclaw state only and does not preserve files elsewhere under /sandbox. Back up paths such as /sandbox/user-data separately before upgrading. For more information, refer to Recover and Rebuild Sandboxes.
  • The starter prompt now loads focused DGX Spark, DGX Station, or Windows WSL Express instructions only after platform detection, keeping unrelated platform guidance out of general onboarding while preserving each platform’s safeguards. The E2E workflow planner also owns typed selector, inference-mode, schema, and Hermes-selection validation before emitting the execution plan.