Connect NVIDIA Free API to Hermes Agent and OpenClaw

Use NVIDIA’s free API to get Hermes Agent running quickly, then extend the same flow to OpenClaw or move to a more stable unified provider later.

This walkthrough shows how to register for NVIDIA Build, create an API key, connect it to Hermes Agent, and validate that the free models really work.

Create the NVIDIA API key

Start from the NVIDIA Build page and follow the account verification flow in order.

Step 1: start with email login

Open NVIDIA Build and click Login. First-time users can begin with their email address.

Step 2: enter the email address

The system decides whether to sign you in or route you into registration.

Step 3: finish registration if the email is new

Complete the registration process first when the address has not been used before.

Step 4: verify the email

Email verification is required before NVIDIA exposes the API features.

Step 5: complete account verification

After login, open the Verify entry and finish the remaining account checks.

Step 6: bind phone number and create the API key

Enter the phone number, finish the code check, open API Keys, and create a new key.

https://integrate.api.nvidia.com/v1

Connect the key inside Hermes

hermes model

Now move back to the terminal and configure Hermes against the NVIDIA endpoint.

Step 7: choose custom endpoints

Inside Hermes model setup, choose custom endpoints instead of a built-in provider.

Step 8: enter Base URL and API key

Paste the NVIDIA Base URL together with the new API key.

Step 9: confirm that the model list loads

If the free model list appears, the endpoint and key are both valid.

Step 10: choose a model and start Hermes TUI

Pick a model from the list, then launch the TUI.

Step 11: send a real test message

A normal reply is the final proof that the connection is working.

Final checklist

NVIDIA account verified

API key copied safely

Base URL set to https://integrate.api.nvidia.com/v1

Hermes can return a real reply

Is the free route enough?

It is usually enough for learning and light use, but NVIDIA rate-limits the free route, so heavier agent workflows may feel slow.

When you need a more stable provider

If the free route becomes too limited, move the same setup flow to a lower-cost unified provider such as SuperAIAPI by registering, comparing model groups, creating a token, and pasting the new Base URL plus API key into Hermes or OpenClaw.