PlayOn

NVIDIA

Sign up for NVIDIA’s API catalog, create a key, and use it as your PlayOn LLM provider.

Create an NVIDIA account and API key

  1. Go to build.nvidia.com (NVIDIA API catalog / NIM).
  2. Sign in or create an NVIDIA account.
  3. Open the API key section (often labeled Get API Key / NGC or account API keys).
  4. Create a key and copy it somewhere safe. You will paste it only into PlayOn Settings on your host.

NVIDIA’s catalog exposes OpenAI-compatible chat endpoints. PlayOn uses the integrate API base URL shown in Settings defaults above.

Models

See Getting started with AI for the full pick list.

  • Minimum: meta/llama-3.1-8b-instruct
  • Recommended: meta/llama-3.3-70b-instruct

Connect in PlayOn

  1. Open your PlayOn admin UI → Settings → In-app agents.
  2. Set Provider to NVIDIA.
  3. Leave the base URL on the preset default unless NVIDIA documents a different endpoint for your account.
  4. Set Model to meta/llama-3.3-70b-instruct (recommended) or meta/llama-3.1-8b-instruct (minimum).
  5. Paste your NVIDIA API key.
  6. Save.

Verify

  1. Open Map / chat.
  2. Ask something that needs a tool, e.g. “List my servers” or “Search the skill catalog for Minecraft”.
  3. If auth fails, re-check the key and that the model id matches the catalog spelling (including org/model slashes).

Tips

  • Tool-calling quality varies by model — if the minimum 8B struggles, use meta/llama-3.3-70b-instruct.
  • Confirm gates still apply for destructive actions in the UI.
  • You can keep NVIDIA for in-app chat and still use MCP with Cursor / OpenClaw / Hermes for the same tools.