# NVIDIA NeMo Agent Toolkit

> Source: https://aiwiki.ai/wiki/nemo_agent_toolkit
> Updated: 2026-07-25
> Categories: AI Agents, Developer Tools, NVIDIA, Open Source AI
> License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/)
> From AI Wiki (https://aiwiki.ai), the free encyclopedia of artificial intelligence. Reuse freely with attribution to "AI Wiki (aiwiki.ai)".

| Field | Value |
| --- | --- |
| Developer | [NVIDIA](/wiki/nvidia) |
| Type | Open-source library for building, profiling, and optimizing multi-agent AI systems |
| Former names | AgentIQ; Agent Intelligence Toolkit (AIQ Toolkit) |
| Initial release | March 2025 (announced at [GTC](/wiki/gtc) as AgentIQ); open-sourced April 2025 |
| Current stable version | 1.8.0 (June 17, 2026); 1.9 alpha/nightly builds also published |
| PyPI package | `nvidia-nat` |
| Written in | Python (3.11 to 3.13) |
| License | Apache 2.0 |
| Repository | github.com/NVIDIA/NeMo-Agent-Toolkit |
| Documentation | docs.nvidia.com/nemo/agent-toolkit |
| Part of | [NVIDIA NeMo](/wiki/nvidia_nemo) suite; NVIDIA AI Enterprise; broader NVIDIA Agent Toolkit |

The **NVIDIA NeMo Agent Toolkit** is an open-source, framework-agnostic library for connecting, profiling, evaluating, and optimizing teams of [AI agents](/wiki/ai_agent). Released by [NVIDIA](/wiki/nvidia) in 2025, first as **AgentIQ** and briefly as the **Agent Intelligence Toolkit** (AIQ Toolkit), it adds a common operational layer (instrumentation, observability, evaluation, optimization, and reusable connectors) on top of [agentic AI](/wiki/agentic_ai) systems without forcing teams to rebuild them on a new platform.[1][7] It is distributed as the PyPI package `nvidia-nat` and belongs to the broader [NVIDIA NeMo](/wiki/nvidia_nemo) software family.[2][4] NVIDIA also places it within the broader **NVIDIA Agent Toolkit**, a collection rather than a single Python package. At SIGGRAPH 2026, NVIDIA added Omniverse libraries and skills to that broader collection; it did not describe them as additions to the `nvidia-nat` 1.8 package.[11][18]

## What is the NeMo Agent Toolkit

The NeMo Agent Toolkit is not a new agent framework meant to replace the ones developers already use. It is a layer that sits alongside them.[16] NVIDIA describes it as an open-source library "for efficiently connecting and optimizing teams of AI agents," and, in its own documentation, as a way to add "intelligence to AI agents across any framework, enhancing speed, accuracy, and decision-making through enterprise-grade instrumentation, observability, and continuous learning."[1][3]

The central design idea is that every agent, tool, and multi-step workflow is treated as a composable function call. Because the unit of composition is a plain function call rather than a framework-specific object, components written for different agent stacks can be mixed, wrapped, and reused inside one system.[1][7] On top of that abstraction the toolkit layers the operational tooling that production agent systems usually lack: end-to-end profiling (down to individual tokens), observability, a repeatable evaluation [harness](/wiki/harness), and automated optimization.[1][7]

In practice this means a team that has already built a multi-agent application (for example a [LangGraph](/wiki/langgraph) research agent calling a [CrewAI](/wiki/crewai) crew and some custom Python tools) can wrap those pieces with the toolkit to measure where time and tokens are spent, trace execution, run evaluations, and then tune the workflow for speed, accuracy, and cost, all without replatforming.[1][9] Early coverage described it as "a lightweight and flexible Python library designed to unify agentic workflows across frameworks, memory systems, and data sources," functioning "as a complementary layer rather than a replacement framework."[7]

## Name history (AgentIQ, Agent Intelligence Toolkit, NeMo Agent Toolkit)

The project has been renamed twice, but NVIDIA has kept the codebase, API, and version numbering continuous across each change, so upgrading has mostly meant changing a dependency name.[2][6] It began life as **AgentIQ**, introduced alongside the open [Llama Nemotron](/wiki/llama_nemotron) reasoning models at NVIDIA's GTC conference in March 2025 and open-sourced in early April 2025.[8][7] It was then renamed the **Agent Intelligence Toolkit** (shortened to AIQ Toolkit), and finally the **NeMo Agent Toolkit**, a change NVIDIA said was made "to better reflect the purpose of the toolkit and to align with the NVIDIA NeMo family of products," with "the core technologies, performance and roadmap" unchanged.[6]

The package renames track the version history recorded in the toolkit's own migration guide:[2]

| Version | Name | PyPI package | Notes |
| --- | --- | --- | --- |
| Initial (March to April 2025) | AgentIQ | `agentiq` | Announced at GTC 2025; open-sourced on GitHub[8][7] |
| v1.1.0 (2025) | Agent Intelligence Toolkit (AIQ Toolkit) | `aiqtoolkit` | `agentiq` renamed to `aiqtoolkit`; API unchanged[2] |
| v1.2.0 (mid-2025) | NeMo Agent Toolkit | `nvidia-nat` | `aiqtoolkit` renamed to `nvidia-nat`; aligned with NeMo family[2][6] |
| v1.5.0 (March 2026) | NeMo Agent Toolkit | `nvidia-nat` | Transitional `aiqtoolkit` package removed[2] |
| v1.8.0 (June 17, 2026) | NeMo Agent Toolkit | `nvidia-nat` | Current stable release; PyPI also carries 1.9 alpha/nightly builds[4][5] |

The acronym NAT in the package name stands for NeMo Agent Toolkit.[1] NVIDIA advised existing users to update their dependency from `agentiq` or `aiqtoolkit` to `nvidia-nat`; compatibility aliases were kept for a few releases and then removed at v1.2.0 and v1.5.0 respectively.[2] The current stable release, v1.8.0, was published on PyPI on June 17, 2026 under the Apache 2.0 license and supports Python 3.11, 3.12, and 3.13. PyPI also lists 1.9 alpha/nightly builds, which are prereleases rather than a newer stable line.[4][5]

## Architecture and how it works

The toolkit is organized around three activities: building agents, gaining insight into them, and optimizing them.[1][3]

**Framework-agnostic wrapping.** The toolkit connects to existing agent frameworks rather than owning the agent logic. It integrates with [LangChain](/wiki/langchain) and LangGraph, [LlamaIndex](/wiki/llamaindex), CrewAI, Microsoft [Semantic Kernel](/wiki/semantic_kernel), Google ADK, and Agno, as well as custom enterprise frameworks and plain Python agents.[1] Because agents and tools are exposed as function calls, the same tool can be reused across frameworks, and reusable connectors let agents reach data sources and services without bespoke glue code.[1][7]

**Profiling.** The toolkit instruments a workflow end to end, tracking token usage, response timings, and hidden latencies at a fine grain. NVIDIA describes the ability to "profile entire workflows from the agent level all the way down to individual tokens," which surfaces the bottlenecks that make multi-agent systems slow or expensive.[1][7]

**Observability.** Telemetry is emitted in an OpenTelemetry-compatible form, so traces and metrics can flow into existing observability platforms (for example [LangSmith](/wiki/langsmith) or other tracing backends) to show how each part of a workflow behaves in production.[1][7]

**Evaluation.** A built-in evaluation system provides a consistent, repeatable way to validate agentic workflows, including [retrieval-augmented generation (RAG)](/wiki/retrieval_augmented_generation) pipelines and full end-to-end runs, so accuracy can be checked as a workflow evolves.[1][7]

**Optimization.** Beyond measurement, the toolkit can tune workflows: hyper-parameter and prompt optimization, [reinforcement-learning](/wiki/reinforcement_learning) [fine-tuning](/wiki/fine_tuning), integration with NVIDIA Dynamo for [inference](/wiki/inference) efficiency, and a set of Agent Performance Primitives (APP) that support parallel execution and priority routing across a multi-agent workflow.[1] A built-in chat UI lets developers interact with a wrapped workflow while these tools run.[1]

**Interoperability protocols.** The toolkit supports the [Model Context Protocol](/wiki/model_context_protocol) (MCP) on both sides: it can act as an MCP client to connect to remote MCP servers, and as an MCP server to publish its own tools to any MCP client.[12] It also works with Agent-to-Agent (A2A) workflows, letting agents built with the toolkit cooperate with agents running elsewhere.[1][12] This is the same interoperability layer that patterns such as [deep agents](/wiki/deep_agents) rely on when they orchestrate long-running, multi-step work.

## Framework integrations

The toolkit is deliberately vendor-neutral about which agent framework a team starts from. The frameworks and protocols documented as first-class integrations include:[1][3]

| Framework or protocol | Role | How the toolkit uses it |
| --- | --- | --- |
| LangChain / LangGraph | Agent and graph orchestration | Automatic wrapping of chains, tools, and graph nodes as function calls |
| LlamaIndex | Data and RAG agents | Wrap retrieval and query engines for profiling and evaluation |
| CrewAI | Multi-agent crews | Wrap crews and roles as composable functions |
| Microsoft Semantic Kernel | Agent framework | Reuse skills and planners inside toolkit workflows |
| Google ADK | Agent Development Kit | Integrate ADK-built agents into a common layer |
| Agno | Agent framework | Wrap Agno agents alongside other frameworks |
| Model Context Protocol (MCP) | Tool and context protocol | MCP client and MCP server support |
| Agent-to-Agent (A2A) | Inter-agent protocol | Interoperate with external agents |
| Plain Python / custom | Any code | Expose arbitrary functions as agent tools |

NVIDIA credits CrewAI, LangChain, LlamaIndex, and Semantic Kernel in the project's acknowledgments as open-source projects the toolkit builds on, and the documentation also lists further connectors and integrations beyond the core frameworks.[4][3] The consistent message is that adoption should be additive: teams keep their framework of choice and gain the operational layer.[1][9]

## The SIGGRAPH 2026 expansion: Omniverse libraries and the local agent stack

At SIGGRAPH 2026, in an announcement dated July 20 and amplified by the supplied NVIDIA Omniverse post, NVIDIA said the **broader NVIDIA Agent Toolkit** now includes [NVIDIA Omniverse](/wiki/nvidia_omniverse) libraries and skills for bringing [physical AI](/wiki/physical_ai) capabilities into existing 3D applications.[11][12][15][17] The name is important. NVIDIA describes NeMo Agent Toolkit, the Apache-licensed `nvidia-nat` library covered by this article, as one component of that broader collection. The SIGGRAPH release does not say that the Omniverse components were merged into the NeMo repository or shipped in version 1.8.[11][18]

The stated goal is to help agents inspect scenes, test changes, and prepare simulation-ready environments for robots, factories, and autonomous systems. NVIDIA's announcement says that simulation content needs more than visual realism: assets also need suitable structure, scale, materials, labels, sensor definitions, and physical properties.[11] OpenUSD supplies the shared scene-description layer, while NVIDIA's SimReady framework defines requirements and validation paths for the physics, semantics, materials, articulation data, and metadata required by a target simulation workflow.[28][29]

The announcement grouped three differently packaged items under the Omniverse-library story. They are publicly discoverable on GitHub, but their code, binaries, maturity, and licenses are not uniform:[11][19][20][21]

| Component | Form and capability | Availability, maturity, and license |
| --- | --- | --- |
| `ovrtx` | C and Python SDK/library that embeds Omniverse RTX rendering and physically based camera, lidar, radar, and other sensor simulation into a host application's own loop; it reads OpenUSD scenes and can return CPU or GPU tensors | Pre-release. The public repository distributes binary packages and examples, requires a compatible NVIDIA RTX GPU and driver for runtime initialization, is closed to external contributions, and places the software under the NVIDIA Software License Agreement and NVIDIA AI Product Specific Terms, not Apache 2.0[19][30] |
| `ovphysx` | Standalone, USD-first wrapper around NVIDIA PhysX with a C API, Python bindings, DLPack tensor exchange, environment cloning for batched [reinforcement learning](/wiki/reinforcement_learning), and CPU or NVIDIA-GPU simulation | Pre-release and not yet mature. Source is BSD-3-Clause, while prebuilt SDK packages and Python wheels use the NVIDIA Omniverse License[20] |
| CAD-to-SimReady | Documentation-driven Agent Skill that coordinates CAD/source conversion, material and physics assignment, SimReady conformance, validation, rendering, and optional packaging; it is not a third compiled library or a monolithic runner | Version 0.1.0 and Apache 2.0. A full property-assignment workflow requires services and dependencies including Docker, NVIDIA Container Toolkit, an NVIDIA GPU, and credentials or existing endpoints; conversion-only work can skip Content Agents[21] |

This distinction also narrows what "openly available" meant in NVIDIA's release. The NeMo Agent Toolkit itself remains open source under Apache 2.0. The CAD Agent Skill and project-authored parts of the Blender example are Apache-licensed, and `ovphysx` source is BSD-licensed. The `ovrtx` SDK and some prebuilt runtimes use NVIDIA licenses. A public GitHub repository is therefore not evidence that every component or binary is open source.[1][19][20][21][22]

NVIDIA showed the components through "SimReady Blender," described in the press release as a sample workflow built with Omniverse libraries and [NemoClaw](/wiki/nvidia_nemoclaw), and released a public Blender integration blueprint/example.[11][12] Its current repository is explicit about status: it is a sample for Blender 5.1 on supported NVIDIA RTX workstations, not a supported production renderer, and it does not reproduce every Cycles, EEVEE, or Blender feature. Its OVPhysX path is a focused SimReady prop check rather than a complete Blender physics replacement, robot-authoring tool, or Isaac Sim workflow.[22]

Adopter language in and around the announcement ranges from launched product use to early exploration:

| Organization | Evidence reviewed | Status supported by the source |
| --- | --- | --- |
| SideFX | SideFX says it is exploring how `ovrtx`, `ovphysx`, OpenUSD, and agent-ready tools could support Houdini workflows | Exploration, not a shipping Houdini integration or production deployment[23] |
| PTC Onshape | The Onshape Render Studio team says it is exploring `ovrtx`, calls the results early, and points users to Onshape Labs early access | Exploration/early access, despite the NVIDIA release's stronger "is using" wording[24] |
| Lightwheel | Lightwheel announced SimReadyGen and says it integrated NVIDIA Omniverse Content Agents with OpenUSD for material assignment, physics classification, texture generation, and validation | First-party product launch and integration claim; performance and physical-accuracy statements remain vendor claims[25] |
| Palatial | NVIDIA says Palatial is using CAD-to-SimReady skills to automate SimReady asset creation and validation at scale | NVIDIA-reported use; no first-party Palatial source was found in this audit[11] |
| ForgeCAD and Moonlake AI | NVIDIA says both are exploring agent-driven 3D content workflows | Exploration[11] |

The companion SIGGRAPH blog presented a separate local stack on a [NVIDIA DGX Station](/wiki/nvidia_dgx_station): NemoClaw, the 550-billion-parameter [NVIDIA Nemotron 3](/wiki/nemotron_3) Ultra model, Omniverse tools and skills, and the OpenShell runtime.[12] NVIDIA claimed that the demonstrated DGX Station configuration could be set up in three steps and roughly 30 minutes and then run with no internet required. That is a vendor statement about the demonstrated local runtime, not an independent benchmark or an air-gap certification.[12]

The maturity of the surrounding software matters too. NemoClaw is an Apache-licensed open-source reference stack in early preview. It orchestrates a selected agent or harness, inference routing, a versioned blueprint, policies, lifecycle tooling, and an OpenShell sandbox.[27] OpenShell is Apache-licensed, but its current repository labels it alpha "single-player mode" for one developer, one environment, and one gateway, with multi-tenant enterprise deployment still a future direction.[26] The press release also said its demo **previewed** local Omniverse workflows from planned RTX Spark systems to orderable GB300-powered DGX Station systems. At the component level, `ovrtx` and the Blender sample require compatible NVIDIA RTX hardware, while `ovphysx` can fall back to CPU simulation.[11][19][20][22]

## Relationship to NeMo and the rest of NVIDIA's agent stack

The 2025 rename folded the toolkit into the NVIDIA NeMo family, the company's broader suite for building, customizing, and operating AI models and agents.[6] By 2026 NVIDIA was also using **NVIDIA Agent Toolkit** as an umbrella for models, agent blueprints, secure runtimes, libraries, and skills. NeMo Agent Toolkit is part of that collection, but it is not the same thing as the whole collection.[18]

- **NeMo Agent Toolkit (this page)** is the Apache-licensed, framework-agnostic Python layer for building, wrapping, profiling, evaluating, and optimizing agents and workflows.[1][4]
- **NVIDIA Agent Toolkit** is the broader product family/collection used in the SIGGRAPH announcement. The Omniverse libraries and skills sit in that umbrella; they are not all modules of `nvidia-nat`.[11][18]
- **NVIDIA NemoClaw** is an Apache-licensed reference stack and blueprint system, not the NeMo profiler. It configures supported agent harnesses inside OpenShell, with inference routing, policies, and lifecycle management. NVIDIA's documentation still labels it early preview.[27]
- **NVIDIA OpenShell** is the sandbox and policy-enforcement runtime underneath NemoClaw. It is open source under Apache 2.0, but its repository currently labels the software alpha and single-user/single-environment rather than production multi-tenant infrastructure.[26]
- **Nemotron 3** and other Nemotron models can supply the reasoning model. NeMo Agent Toolkit is model-agnostic, while NemoClaw can route to local or hosted inference providers.[12][27]
- **[NVIDIA NIM](/wiki/nvidia_nim)** microservices provide packaged model inference. NeMo Agent Toolkit and NIM were co-integrated into Microsoft Azure AI Foundry under the project's earlier AgentIQ name.[9]

NeMo Agent Toolkit is also distributed through NVIDIA AI Enterprise. That positions it between agent frameworks and NVIDIA's model/inference stack, but it does not give all adjacent NVIDIA Agent Toolkit components the NeMo library's Apache license or release status.[1][11][18]

## Adoption and use cases

Because the toolkit is additive rather than a replacement, its adoption has largely come through integrations with existing platforms. When AgentIQ launched at GTC 2025, NVIDIA and Microsoft integrated it, together with NVIDIA NIM microservices, into Microsoft Azure AI Foundry; Microsoft framed the combination as a way to speed up enterprise agent projects that can otherwise run 9 to 12 months from start to production.[9][10] It is also distributed inside NVIDIA AI Enterprise, which is how many organizations encounter it.[1]

Typical use cases fall into a few buckets:

- **Connecting and reusing agents and tools** across frameworks so an enterprise does not have to standardize on one agent library.[1][7]
- **Profiling and cost control** for multi-agent systems, identifying the tokens, calls, and latencies that dominate a workflow's time and spend.[1][7]
- **Evaluation and continuous improvement**, running repeatable checks on RAG and end-to-end agent workflows as they change.[7]
- **Simulation and 3D content through the broader NVIDIA Agent Toolkit**, where adjacent Omniverse components can expose sensor simulation, physics, and SimReady preparation to agents. The reviewed SideFX and PTC work is exploratory, the public Blender integration is a sample, and Lightwheel's SimReadyGen is the clearest first-party launched integration among the organizations named in the July release.[11][22][23][24][25]
- **Local/offline agent stacks**, such as NVIDIA's DGX Station demonstration combining NemoClaw, Nemotron 3 Ultra, Omniverse tools, and OpenShell. NVIDIA said that demonstrated runtime needed no internet connection, but did not characterize it as a certified air-gapped deployment.[12][26][27]

## Significance

The toolkit's significance is less about any single feature and more about where it sits. The 2025 to 2026 wave of agentic AI produced many competing frameworks (LangChain and LangGraph, CrewAI, LlamaIndex, Semantic Kernel, Google ADK, and others), and enterprises frequently ended up with agents spread across several of them. By treating agents and tools as function calls and adding a shared layer for profiling, observability, evaluation, and optimization, the toolkit offers an interoperability and operations tier that does not force a team to pick one framework or rebuild.[1][7] That "wrap what you have" posture, plus an Apache 2.0 license and OpenTelemetry-compatible telemetry, made it an easy addition to existing stacks.[1]

The SIGGRAPH 2026 announcement is significant as ecosystem context, not as a NeMo 1.8 feature release. It shows NVIDIA expanding the broader Agent Toolkit concept beyond text agents into physical-AI tools, OpenUSD scene data, sensor simulation, physics, and SimReady asset preparation. The public artifacts also show the limits of the label: some are open-source skills or source code, some are proprietary pre-release binaries, and some are demonstrations or exploratory partner work.[11][19][20][21][22]

## Limitations

- **The core license does not cover the whole NVIDIA Agent Toolkit umbrella.** `nvidia-nat` is Apache 2.0, but the adjacent SIGGRAPH components use mixed terms. `ovrtx` is pre-release software under NVIDIA's software license; `ovphysx` combines BSD-licensed source with NVIDIA-licensed prebuilt packages; CAD-to-SimReady is an Apache-licensed Agent Skill; and the Blender sample preserves the separate licenses of its runtimes.[4][19][20][21][22] Secondary coverage that equates public GitHub access with a single open-source license misses this distinction.[13]
- **An operational layer, not a complete autonomous-agent runtime.** NeMo Agent Toolkit can build and wrap workflows, but its defining strengths are cross-framework composition, profiling, evaluation, and optimization. Sandboxing, a chosen agent harness, model inference, and domain tools come from other components.[1][7][18]
- **Governance of agent-produced work.** Independent analysis has raised the question of who governs the code and outputs that toolkit-optimized agents generate, a concern that grows as agents take on more autonomous engineering tasks.[14]
- **Several adjacent SIGGRAPH components remain pre-release or sample-grade.** `ovrtx` and `ovphysx` are pre-release, the Blender repository disclaims production-renderer status, NemoClaw is early preview, and OpenShell describes itself as alpha single-player software. Those labels do not establish production reliability or multi-tenant readiness.[19][20][22][26][27]
- **Hardware and offline claims are narrower than a general local-support claim.** `ovrtx` and the Blender sample require compatible NVIDIA RTX hardware; `ovphysx` alone can use CPU simulation. NVIDIA's no-internet, three-step, roughly 30-minute statement applies to its demonstrated DGX Station stack, and the RTX Spark-to-DGX range was presented as a preview.[11][12][19][20][22]
- **Named adopters are at different stages.** Lightwheel announced an integration, NVIDIA reports Palatial use, and SideFX, PTC, ForgeCAD, and Moonlake are described as exploring or in early access. The announcement is not evidence that all named organizations operate these workflows in production.[11][23][24][25]
- **Vendor benchmark and capability claims.** Figures such as the Nemotron 3 Ultra parameter count and the DGX Station setup time are NVIDIA's own statements and should be read as vendor claims rather than independently verified results.[12]

## In plain terms (ELI5)

Imagine several AI assistants working together on a job. One is good at research, one at writing, one at pulling data, and each was built with different tools, almost like workers who each speak a different language. The NeMo Agent Toolkit is like a universal adapter plus a stopwatch plus a supervisor. The adapter lets the assistants plug into one another even though they were built differently. The stopwatch measures exactly how long each step takes and how much it costs, so you can find the slow, expensive parts. The supervisor then helps them work faster and more accurately. The important part is that nobody has to quit their job and start over: the toolkit wraps around the team you already have.

The 2026 SIGGRAPH news concerned a larger toolbox with a similar name. NVIDIA added separate 3D tools and skills to the broader NVIDIA Agent Toolkit so agents could help prepare simulated worlds. Those items can work beside NeMo Agent Toolkit, but they are not all inside its Python package, and they do not all share its open-source license or production maturity.

## See also

- [NVIDIA NemoClaw](/wiki/nvidia_nemoclaw)
- [NVIDIA NeMo](/wiki/nvidia_nemo)
- [Nemotron 3](/wiki/nemotron_3)
- [NVIDIA Omniverse](/wiki/nvidia_omniverse)
- [Model Context Protocol](/wiki/model_context_protocol)
- [Agentic AI](/wiki/agentic_ai)
- [LangGraph](/wiki/langgraph)
- [NVIDIA DGX Station](/wiki/nvidia_dgx_station)

## References

1. NVIDIA, "NeMo-Agent-Toolkit (README)," GitHub. https://github.com/NVIDIA/NeMo-Agent-Toolkit
2. NVIDIA, "Migration Guide," NeMo Agent Toolkit documentation, GitHub. https://github.com/NVIDIA/NeMo-Agent-Toolkit/blob/develop/docs/source/resources/migration-guide.md
3. NVIDIA, "NVIDIA NeMo Agent Toolkit Overview," NVIDIA Developer. https://developer.nvidia.com/nemo-agent-toolkit
4. "nvidia-nat," Python Package Index (PyPI). https://pypi.org/project/nvidia-nat/
5. NVIDIA, "Releases, NVIDIA/NeMo-Agent-Toolkit," GitHub. https://github.com/NVIDIA/NeMo-Agent-Toolkit/releases
6. NVIDIA, "NVIDIA Agent Intelligence toolkit is now the NVIDIA NeMo Agent toolkit," NVIDIA Developer Forums, 2025. https://forums.developer.nvidia.com/t/nvidia-agent-intelligence-toolkit-is-now-the-nvidia-nemo-agent-toolkit/335881
7. MarkTechPost, "NVIDIA AI Released AgentIQ: An Open-Source Library for Efficiently Connecting and Optimizing Teams of AI Agents," April 5, 2025. https://www.marktechpost.com/2025/04/05/nvidia-ai-released-agentiq-an-open-source-library-for-efficiently-connecting-and-optimizing-teams-of-ai-agents/
8. NVIDIA, "NVIDIA Launches Family of Open Reasoning AI Models for Developers and Enterprises to Build Agentic AI Platforms," NVIDIA Newsroom, March 18, 2025. https://nvidianews.nvidia.com/news/nvidia-launches-family-of-open-reasoning-ai-models-for-developers-and-enterprises-to-build-agentic-ai-platforms
9. Microsoft, "Accelerating Agentic Workflows with NVIDIA AgentIQ, Azure AI Foundry and Semantic Kernel," Microsoft Agent Framework, March 18, 2025. https://devblogs.microsoft.com/agent-framework/accelerating-agentic-workflows-with-nvidia-agentiq-azure-ai-foundry-and-semantic-kernel/
10. InfoQ, "Azure AI Foundry Integrates NVIDIA NIM and AgentIQ," March 2025. https://www.infoq.com/news/2025/03/azure-foundry-nim-agent-iq/
11. NVIDIA, "NVIDIA Agent Toolkit Expands With New Omniverse Libraries, Putting AI Agents to Work Building Simulation-Ready Worlds," NVIDIA Newsroom, July 20, 2026. https://nvidianews.nvidia.com/news/nvidia-agent-toolkit-expands-with-new-omniverse-libraries-putting-ai-agents-to-work-building-simulation-ready-worlds
12. NVIDIA, "At SIGGRAPH, NVIDIA Advances Graphics and Simulation With Agentic and Physical AI," NVIDIA Blog, July 20, 2026. https://blogs.nvidia.com/blog/siggraph-news-2026/
13. UC Today, "NVIDIA Omniverse Libraries: Open Source, But There's a Catch," July 2026. https://www.uctoday.com/immersive-workplace-xr-tech/nvidia-omniverse-libraries-agent-toolkit-siggraph-2026/
14. Mneme HQ, "NVIDIA NeMo Agent Toolkit Optimizes AI Agent Workflows. Who Governs the Code They Produce?" https://mnemehq.com/insights/nvidia-nemo-agent-toolkit-engineering-governance/
15. NVIDIA, "NVIDIA Agent Toolkit Expands With New Omniverse Libraries" (press release full text), GlobeNewswire, July 20, 2026. https://www.globenewswire.com/news-release/2026/07/20/3329895/0/en/NVIDIA-Agent-Toolkit-Expands-With-New-Omniverse-Libraries-Putting-AI-Agents-to-Work-Building-Simulation-Ready-Worlds.html
16. NVIDIA, "NVIDIA NeMo Agent Toolkit Overview (1.8)," docs.nvidia.com. https://docs.nvidia.com/nemo/agent-toolkit/1.8/index.html
17. NVIDIA Omniverse, "NVIDIA Agent Toolkit Expands With New Omniverse Libraries," X post, July 2026. https://x.com/nvidiaomniverse/status/2079223282759282949
18. NVIDIA, "NVIDIA NeMo Agent Toolkit," NVIDIA Developer. https://developer.nvidia.com/nemo-agent-toolkit
19. NVIDIA Omniverse, "ovrtx," GitHub. https://github.com/NVIDIA-Omniverse/ovrtx
20. NVIDIA Omniverse, "ovphysx," NVIDIA PhysX repository, GitHub. https://github.com/NVIDIA-Omniverse/PhysX/tree/main/ovphysx
21. NVIDIA, "omniverse-cad-to-simready Agent Skill," GitHub. https://github.com/NVIDIA/skills/tree/main/skills/omniverse-cad-to-simready
22. NVIDIA Omniverse, "ovrtx Blender Example," GitHub. https://github.com/NVIDIA-Omniverse/omniverse-labs/tree/main/projects/ov-blender-example
23. SideFX, "SideFX Explores NVIDIA Omniverse Libraries for Houdini," July 20, 2026. https://www.sidefx.com/community/sidefx-explores-nvidia-omniverse-libraries-for-houdini/
24. PTC Onshape, "Exploring NVIDIA Omniverse Library Integration within Onshape's Cloud-Native Design Workflow," July 20, 2026. https://www.onshape.com/en/blog/cloud-native-cad-render-studio-nvidia-ovrtx
25. Lightwheel, "Introducing SimReadyGen," July 20, 2026. https://lightwheel.ai/media/simreadygen
26. NVIDIA, "OpenShell," GitHub. https://github.com/NVIDIA/OpenShell
27. NVIDIA, "Overview of NVIDIA NemoClaw" and "Release Notes," NVIDIA documentation. https://docs.nvidia.com/nemoclaw/latest/about/overview.html and https://docs.nvidia.com/nemoclaw/latest/user-guide/openclaw/release-notes
28. NVIDIA, "What Is OpenUSD?" NVIDIA Glossary. https://www.nvidia.com/en-us/glossary/openusd/
29. NVIDIA, "What Is SimReady?" NVIDIA Glossary. https://www.nvidia.com/en-us/glossary/simready/
30. NVIDIA, "Integrate NVIDIA Omniverse RTX Sensor Simulation Into Existing Apps," NVIDIA Technical Blog, July 2026. https://developer.nvidia.com/blog/integrate-nvidia-omniverse-rtx-sensor-simulation-into-existing-apps/