Release Notes

July 9, 2026

v0.0.79

NemoClaw v0.0.79 expands hosted and local inference options, improves operator diagnostics and shell integration, and hardens sandbox recovery, Deep Agents runtime limits, policy boundaries, and release validation. The release also refreshes quickstarts and variant rendering so OpenClaw, Hermes, and LangChain Deep Agents Code readers see only the agent-specific setup paths that apply to them.

  • OpenRouter is now a first-class onboarding provider for OpenClaw, Hermes, and LangChain Deep Agents Code. NemoClaw validates sk-or- credentials, uses OpenRouter’s OpenAI-compatible route through https://inference.local/v1, and keeps the OpenRouter catalog independent from NVIDIA Endpoints filtering. For more information, refer to Choose an Inference Provider, Switch Inference Providers, and Platform Support and Launch Claims.
  • Local and compatible inference setup is more resilient. Managed vLLM starts with argv-safe Docker arguments, DGX Spark uses the Qwen3.6 qwen3_coder tool parser, OpenClaw bounds managed inference compaction, and onboarding uses calibrated provider-validation and gateway-readiness deadlines to avoid hanging or failing too early on slower hosts. For more information, refer to Choose an Inference Provider, Choose a Local Inference Server, and Troubleshooting.
  • CLI and installer diagnostics are easier to act on. The new nemoclaw completion command generates Bash, Zsh, and Fish completions, structured logging now supports --debug and NEMOCLAW_LOG_LEVEL, status warns when the live gateway route drifts from the recorded sandbox route, and gateway teardown now recommends and runs openshell gateway remove consistently across platforms. Installer and uninstall flows also report preserved orphaned sandboxes more honestly, and DGX Spark express installs use the expected my-assistant sandbox name. For more information, refer to NemoClaw CLI Commands Reference, Switch Inference Providers, and Host Files and State.
  • Runtime and configuration boundaries are tighter. Managed LangChain Deep Agents Code enforces process and file-descriptor limits, accepted OTLP endpoint URLs no longer trip the secret guard, proxy environment exports stay POSIX-compatible, nested OpenShell PID namespaces are recognized by config guards, and TOML config reads and parsing fail closed instead of accepting ambiguous content. Re-onboarding also removes stale registered extra providers so the sandbox state matches the selected provider set. For more information, refer to Security Best Practices, Credential Storage, NemoClaw CLI Commands Reference, and Quickstart with LangChain Deep Agents Code.
  • Network policy and messaging behavior now preserve narrower boundaries. Homebrew Git operations require the GitHub policy preset for Git egress, Gmail has a documented policy preset and integration example, custom URL-based MCP server allowlists are documented, WhatsApp post-pair gateway restarts no longer require operator.admin, and the overview hides messaging-channel guidance from Deep Agents pages where the channel flow is not supported. For more information, refer to Network Policies, Common Integration Policy Examples, Customize the Network Policy, and Choose Messaging Channels.
  • Onboarding and recovery paths preserve intent across more interrupted flows. Resume recovery now follows one explicit finite-state path, pending route reservations survive resume, sandbox create-failure reporting is separated from create-step handling, BuildKit progress no longer forces plain output, and null-name resume sessions are covered so canceled or malformed session state does not send users down the wrong recovery path. For more information, refer to NemoClaw Quickstart with OpenClaw, NemoClaw CLI Commands Reference, Recover and Rebuild Sandboxes, and Troubleshooting.
  • Documentation and release validation are more deterministic. The docs define extension taxonomy and SDK readiness gates, streamline the agent quickstarts, clarify legacy k3s sandbox resources, preserve list spacing in generated agent-variant pages, and add release-train risk planning with a PR E2E check, queued Jetson dispatch guards, TUI idle regression coverage, and reusable live-readiness polling primitives. For more information, refer to Extension Taxonomy and SDK Readiness, NemoClaw Quickstart with OpenClaw, Architecture Details, and the NemoClaw E2E README.