At a glance
- Availability: Experimental (how to enable).
- Authentication: API key.
- Connection: The key comes from
HF_TOKEN. - Provider documentation: Authentication reference.
Credentials
Set these per environment. See Connect an integration.Setup
- Create a Hugging Face account: Go to https://huggingface.co and sign up or sign in. Hub API access and a generous Inference Providers free tier are available without billing.
- Create an access token: Open https://huggingface.co/settings/tokens and create a fine-grained token. Enable ‘Make calls to Inference Providers’ if you want to use the chat completion tool; read access covers the Hub search tools.
- Store the token: Copy the token and add it to your .env file as HF_TOKEN=hf_…
- Verify access: Run the Who Am I tool to confirm the token works and shows the expected permissions.
Provider notes
- Hub search endpoints work with read-scoped tokens; chat completions require the Inference Providers permission
- Inference Providers usage beyond the free tier is billed at provider rates with no markup; PRO accounts include monthly credits
- Some gated models (e.g. meta-llama) require accepting their license on the Hub before access
Tools
Verify the connection
Call a read tool such ashuggingface__search_models with arguments for your account. Confirm that the result comes from the intended account or workspace before enabling write tools.