About | NVIDIA NeMo Relay

Use Getting Started to choose the smallest NeMo Relay path that matches the boundary you want to observe or control. Relay can run beside a local coding-agent harness, wrap application-owned tool or LLM calls, connect through a maintained framework integration, or activate reusable plugin behavior from configuration.

Using an AI Coding Assistant?

Install NVIDIA agent skills to give compatible coding agents product-specific Relay workflows and references. After the NeMo Relay skills are published in nvidia/skills, install the full Relay skill set without cloning the catalog:

npx skills add nvidia/skills \
  --skill nemo-relay-install \
  --skill nemo-relay-get-started \
  --skill nemo-relay-instrument-calls \
  --skill nemo-relay-instrument-context-isolation \
  --skill nemo-relay-instrument-typed-wrappers \
  --skill nemo-relay-plugin-observability \
  --skill nemo-relay-plugin-adaptive-tuning \
  --skill nemo-relay-plugin-build \
  --skill nemo-relay-migrate-from-flow \
  --skill nemo-relay-debug-runtime-integration \
  --agent codex

Use --agent claude-code, --agent cursor, or another supported agent target when that is where you want the skill installed. To inspect the current NVIDIA skill catalog first, run:

npx skills add nvidia/skills --list

Agent skills help your coding assistant apply Relay documentation and workflows. They do not replace installing NeMo Relay packages, configuring runtime plugins, or validating emitted events.

Find the Right Starting Point

Start with the fastest path that can prove Relay is working, then move toward more involved integration and development workflows only when you need them.

Try now with the CLI: Use NeMo Relay CLI when a local Claude Code or Codex coding-agent harness owns invocation. Relay can observe hooks, gateway-routed model traffic, and exporter output without changing application code.
Use Hermes Agent’s native integration: NeMo Relay is built into Hermes Agent, and Hermes Agent understands NeMo Relay plugin configurations. No separate Relay installation, observability plugin, or Relay CLI setup is required.
Use a supported integration: Start with Supported Integrations when LangChain, LangGraph, Deep Agents, or OpenClaw already owns scheduling, callbacks, tool execution, provider payloads, or agent lifecycle hooks.
Instrument application-owned calls: Start with Quick Start, then use Instrument Applications when your Python, Node.js, or Rust application owns the tool or LLM call site.
Build a custom framework or provider integration: Use Integrate into Frameworks when a new framework, orchestration layer, SDK, or provider adapter owns scheduling, retries, callbacks, and provider payloads.
Package reusable runtime behavior: Use Build Plugins for shared middleware, subscribers, exporters, model pricing, policy, or adaptive behavior. Use Observability for export setup and Adaptive after baseline instrumentation is working.

Getting Started Process

Use the pages in this section in order when you are setting up Relay for the first time:

Review Prerequisites for language, package, and repository tooling.
Use Installation to install the CLI, language binding, or integration package that matches your boundary.
Use Configuration when you need runtime, plugin, observability, or adaptive setup guidance.
Use Quick Start to run the smallest working path and confirm Relay emits the expected output.

If you need the conceptual model before setup, read Agent Runtime Primer and Concepts.