Skip to content

v0.1.0 — 2026-05-20

First public release of NeMo Platform. This release establishes the local-first OSS distribution and the core agent build, optimize, and secure flow.

Highlights

  • Agent lifecycle managed end-to-end through nemo agents (create, deploy, invoke, evaluate, optimize, secure, undeploy).
  • Optimize and secure deployed agents with sibling-candidate validation before promotion.
  • Provider-agnostic Inference Gateway with virtual models and middleware (Switchyard routing, guardrails).
  • Guardrails, vulnerability scanning, and evaluation available out of the box.
  • Local-first install — make bootstrap + nemo setup, no cluster required for this release. Cluster and scale-out deployment are on the roadmap.

What's included

Agents

  • nemo agents lifecycle commands (create, deploy, invoke, evaluate, optimize, undeploy, delete).
  • NAT (NeMo Agent Toolkit) workflows as the agent format.
  • Sibling-agent candidate flow for applying optimization and security changes without editing a running agent in place.

Models and Inference

  • Inference Gateway: provider registration, virtual models, entity-name routing.
  • Switchyard middleware: random routing for strong/weak splits, cross-format translation.
  • Supported upstream providers: NVIDIA Build, OpenAI, Anthropic, Google Gemini, Ollama, generic OpenAI-compatible endpoints.

Guardrails

  • nemo-guardrails middleware applied via virtual models.
  • Catalog content-safety, topic-control, and safety-guard models.

Evaluation

  • LLM-as-a-judge, similarity, RAG, and agentic metrics.
  • Agent and model configuration as evaluation targets.

Vulnerability Scanning

  • garak-backed audits against models reachable through the Inference Gateway.
  • Audit configs, probes, targets, and JSONL and HTML reports.

Data

  • Data Designer for synthetic dataset generation.
  • Anonymizer for PII handling in datasets.

Studio

  • Workspace browser, agent list, and chat playground for deployed agents and models. The Studio agent build, optimize, and secure flow is in progress and not part of 0.1.0.

Coding-agent skills

  • nemo skills install --agent <claude|codex|cursor|opencode> installs platform skills for the supported coding agents.
  • Agent-flow skills: nemo-setup, nemo-explore, nemo-spec, nemo-build-agent, nemo-try-agent, agents-optimize, agents-secure, nemo-status, nemo-teardown.

CLI and SDK

  • nemo CLI with auto-generated API commands plus use-case commands like chat.
  • nemo-platform Python SDK, generated by Stainless from the OpenAPI spec.

Install

git clone https://github.com/NVIDIA-NeMo/nemo-platform.git
cd nemo-platform
make bootstrap
nemo setup

See Setup for prerequisites and provider configuration.

Compatibility

  • Python 3.11–3.13
  • macOS, Linux
  • 16 GB disk, 8 GB RAM
  • Node 22.18.0+ for Studio assets

Current constraints

  • Local-only. Cluster and scale-out deployment are on the roadmap for enterprise use; 0.1.0 supports a single local install.
  • Studio agent flow. Studio ships with chat and workspace browsing today. The agent build, optimize, and security flows in Studio are in progress.
  • In-place agent edits. The Agents v2 API has no patch operation. Config changes go through a sibling agent and redeploy.
  • Bring your own agent. The platform's deploy and optimize paths target NAT workflows. NAT ships with a thin LangGraph wrapper, so LangGraph agents work first-class today. Agents from other frameworks need a thin wrapper to plug into NAT. Broader framework support is on the roadmap.

Coming soon

  • Fine-tuning (Customizer): model fine-tuning workflows on the platform.
  • Safe Synthesizer: privacy-preserving synthetic data generation with differential-privacy guarantees.
  • Studio agent flow: the build, optimize, and secure paths surfaced in the web UI.
  • Cluster and scale-out deployment for enterprise workloads.