Skip to content
 
 

Repository files navigation

NemoClaw Community

License Security Policy

NemoClaw Community is a collection of examples that showcase NemoClaw blueprints for constrained, inspectable agent workflows.

NemoClaw is the blueprint layer for composing three things into a repeatable agent system:

  • Model — the inference endpoint, model selection, and provider configuration the agent uses.
  • Harness — the agent runtime, skills, bridges, state, and workflow-specific behavior.
  • OpenShell — the sandbox, gateway, policy, provider, and networking substrate that runs the harness with explicit boundaries.

The examples in this repository demonstrate complete blueprint patterns: they show how a model is wired to a harness, how the harness is packaged with skills and integrations, and how OpenShell constrains and runs the resulting agent.

Reference Examples

Some examples are included in this repository. Others currently live in brevdev/nemoclaw-demos and are candidates for future consolidation here.

Example Description Link
Personal Community Sentiment Triage Pairs a Hermes harness with an OpenShell sandbox and community-signal integrations across Slack, Outlook, live read-only GitHub REST, GitHub discussion mirrors, and NVIDIA forum mirrors. Guide
Hermes Brev Launchable Provides a notebook path from a fresh Brev CPU instance to a working NemoClaw-managed Hermes sandbox, including installation, onboarding, API verification, and terminal access. Guide
OpenClaw Omni Example Sets up a NemoClaw sandbox with a Nemotron Omni vision sub-agent, including reference OpenClaw configuration, policy, agent instructions, and verification scripts. Guide
Hermes Omni Example Builds a local multimodal Hermes agent that can inspect video, audio, images, and PDFs through Nemotron Omni while running inside an OpenShell-constrained sandbox. Guide
Flight Tracking Example Adds a live airspace console with real-time aircraft data, map controls, aviation overlays, and an agent skill that operates through host-side proxies. Guide
Google Workspace Integration Connects a NemoClaw agent to Gmail, Calendar, Drive, Docs, Sheets, Contacts, and Tasks through a host-side credential flow and sandbox-deployed tools. Guide
Planet Integration Gives the agent read-only access to Planet imagery workflows, including catalog search, thumbnails, tasking estimates, pass availability, and account quota checks. Guide
Wakeup Example Adds a host-controlled schedule that wakes the sandboxed agent at fixed intervals to follow a task list without letting the agent control its own timer. Guide

Getting Started

Choose an example from the table above and follow its guide. To run an example from this repository, clone the repo first:

git clone https://github.com/NVIDIA/nemoclaw-community.git
cd nemoclaw-community

For other examples, follow the linked guide in brevdev/nemoclaw-demos. Each example documents its own host requirements, credentials, setup steps, and OpenShell policy details.

Requirements

  • Linux host with Docker or a compatible container runtime
  • OpenShell CLI and gateway
  • Access to an OpenAI-compatible inference endpoint
  • Optional integration credentials for Slack, Microsoft Graph/Outlook, GitHub live reads, and source ETL mirrors

This project will download and install additional third-party open source software projects. Review the license terms of these open source projects before use. See THIRD-PARTY-NOTICES for the repository inventory.

Contributing

See CONTRIBUTING.md. This project uses Developer Certificate of Origin sign-offs for inbound contributions.

Security

See SECURITY.md. Do not file public GitHub issues for security vulnerabilities.

Support

See SUPPORT.md for support channels and expectations.

Governance And Maintainers

License

This project is licensed under the Apache 2.0 License. See LICENSE for details.

About

Our repository of community driven examples, showcases, and integrations

Resources

Code of conduct

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages