NeMo Platform

Make the agents you ship faster, more accurate, and safer.

NeMo Platform brings NVIDIA NeMo libraries together under one CLI, Python SDK, and web UI. Hardening, evaluation, and tuning for the agents you put in production.

Get started

Prerequisites: Python 3.12-3.13, uv, and an API key for an inference provider (NVIDIA Build, OpenAI, Anthropic, Google Gemini, or a local Ollama instance). Source development needs Git, GNU Make, a C compiler, and either Flox (recommended) or a system toolchain matching make toolchain-versions. Docker is required when starting local services.

Quick install from PyPI:

curl -LsSf https://astral.sh/uv/0.9.30/install.sh | sh
export PATH="$HOME/.local/bin:$PATH"
uv tool install "nemo-platform[all]"

nemo setup

uv tool install gives you a global nemo command in its own isolated environment, with nothing to activate. The all extra adds the platform services, so nemo services run works; without it you get the SDK and CLI only. To import the SDK from your own code, uv pip install "nemo-platform[all]" into a virtual environment instead.

Source checkout for development:

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

# Install Flox first: https://flox.dev/docs/install-flox/install
make bootstrap
flox -q activate

nemo setup

make bootstrap uses the Flox-pinned uv, Node.js, and pnpm toolchain; it does not require a prior flox -q activate. Activate Flox after bootstrap to continue development in the managed environment. Without Flox, install the versions printed by make toolchain-versions and a C compiler, then run make TOOLCHAIN=system bootstrap followed by source .venv/bin/activate. See SETUP.md.

nemo setup starts local services, registers your LLM provider, discovers available models, selects default and fast agent models, installs agent skills, and deploys a sample agent (see more below).

Review Telemetry and Privacy for the omnibus disclosure covering anonymous telemetry, bundled library telemetry, third-party endpoint notes, and opt-out controls.

See SETUP.md for the full source setup playbook (local data dir, DB reset, manual service start, troubleshooting).

Verify:

nemo services status

To permanently reset local state, follow the explicitly confirmed, guarded sequence in SETUP.md. It removes the managed ClickHouse container before deleting any bind-mounted data.

nemo --help                # All commands
nemo models list           # Available models
nemo chat <model-name>     # Chat directly with a model
nemo services status       # Platform health
nemo skills list           # Skills installed on the platform

http://localhost:8080/apis/inference-gateway/v2/workspaces/default/openai/-/v1/chat/completions

nemo services run

export NVIDIA_API_KEY=nvapi...
export NEMO_DEFAULT_MODEL=nvidia-nemotron-3-super-120b-a12b
export NEMO_FAST_MODEL="$NEMO_DEFAULT_MODEL"
nemo setup --auto --start-services --install-skills --deploy-agent

Use NeMo Platform from your coding agent

After installation, launch your coding agent (Claude Code, Codex, Cursor, OpenCode, etc) from inside the nemo-platform directory. This is the primary way of interacting with the NeMo Platform.

Things you can ask it to do, once the platform is running:

• "Scaffold an agent from this spec and deploy it."
• "Run an evaluation against my agent."
• "Add content-safety guardrails to my agent."
• "Help me optimize my agent."
• "Show me what's running on the platform."
• "Shut down NeMo cleanly."

What's here today

• Secure agents. Guardrails (content safety, jailbreak detection, PII redaction), Auditor (red-teaming via garak), Anonymizer (PII handling for training data).
• Evaluate agents. LLM-as-judge, deterministic, agentic, and RAG benchmarks. Harbor-backed eval suites for regression testing.
• Tune agents. Skill optimization, prompt and hyperparameter tuning, Switchyard model routing.
• Build agents. NVIDIA NeMo Agent Toolkit (NAT) for LangGraph-based agents. Shared infrastructure: Inference Gateway, Secrets, Files, Entity Store, Jobs.
• Generate synthetic data. Generate synthetic data for training or evaluation purposes using Data Designer.
• NeMo Studio (alpha). Installed automatically with the platform. Browser UI for chat, monitoring, and reviewing optimization suggestions. Studio's agent-focused features are still a work in progress; the CLI is the primary surface today.

Release notes

See the current release notes for the latest features, improvements, and known limitations.

Skills

nemo setup detects Claude Code, Cursor, Codex, and OpenCode and installs NeMo skills into your agent of choice, either into the local directory or globally. Platform-level skills live under packages/nemo_platform_ext/src/nemo_platform_ext/skills/ and ship with the nemo-platform package; plugin-owned skills live under plugins/<plugin>/src/<plugin>/skills/.

To install or refresh skills:

nemo skills install --agent claude
nemo skills install --agent claude --skill nemo-build-agent --skill nemo-status

Try the demo agent

nemo setup --deploy-agent deploys a demo calculator agent you can use to explore the platform's evaluate / optimize loop.

nemo agents invoke --agent calculator-agent --input "what is 12 * 8?"

The calculator-agent package is installed automatically (plugins/nemo-agents/examples/calculator-agent/).

The demo agent uses ${NEMO_DEFAULT_MODEL} for both execution and the judge LLM. To select different models for either/both, update the yaml config files.

Documentation

Full documentation: NeMo Platform docs

• Telemetry and privacy: anonymous telemetry, data collection, and opt-out controls.
• Setup: installation, providers, SDK.
• CLI reference: all commands.
• API reference: REST endpoints.

Development

See CONTRIBUTING.md for development workflow. See TESTING.md for testing strategy.

License

NeMo Platform is licensed under the Apache License 2.0. Third-party open-source dependencies have their own licenses; review them before use.