Summary

OpenClaw should query the public NVIDIA Catalog API during onboarding

Problem to solve

When users onboard to OpenClaw, we want to surface our best models from build.nvidia.com immediately, without requiring a code update every time NVIDIA launches a new flagship. Currently, model promotion is static/hardcoded.

Proposed solution

OpenClaw should query the public NVIDIA Catalog API on startup:

https://assets.ngc.nvidia.com/products/api-catalog/featured-models.json

This endpoint returns a ranked list of NVIDIA's best models with metadata (model, model-name, context, max-output). OpenClaw should:

1. On startup, fetch this catalog and cache it (re-fetch periodically, e.g. every 24h or on startup)
2. Surface models in onboarding — promote the top-ranked model(s) to users
3. Be futureproof — when NVIDIA adds new models (e.g. the upcoming Omni model launching April 28th), they automatically appear in OpenClaw without any code change

Alternatives considered

Leave list hard-coded, requiring extra openclaw updates.

Impact

Affected users/systems/channels

• New users onboarding to OpenClaw for the first time
• Provider plugin for NVIDIA (build.nvidia.com) integrations
• All channels (Discord, Telegram, etc.) where users interact with the NVIDIA provider

Severity

• Annoying to blocking for new users: they immediately encounter a suboptimal experience and must manually select the best model, creating friction at the worst possible moment.

Frequency

• Always for every new NVIDIA user. Not an edge case.

Consequence

• New users get a generic or outdated model as their first impression of OpenClaw + NVIDIA's best hardware.

Urgency

• Low until new flagship models arrive.

Evidence/examples

No response

Additional information

No response