> ## Documentation Index
> Fetch the complete documentation index at: https://noradocs.solomontsao.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Huggingface

# Hugging Face

> Where to issue a Hugging Face access token and connect it to Nora.

Hugging Face integrations let your agents query the Inference API, push to model/dataset repos, and fetch private model weights.

## Where to apply for credentials

<Card title="Hugging Face Access Tokens" href="https://huggingface.co/settings/tokens" icon="brain" arrow>
  [https://huggingface.co/settings/tokens](https://huggingface.co/settings/tokens)
</Card>

<Steps>
  <Step title="Sign in">
    Open [huggingface.co](https://huggingface.co) and sign in.
  </Step>

  <Step title="Create a token">
    Settings → Access Tokens → **New token**. Pick: - **Read** — for inference API calls + downloading
    public/private models you have access to. - **Write** — additionally allows pushing to your repos.
  </Step>

  <Step title="Copy the token">
    Starts with `hf_...`.
  </Step>
</Steps>

## Connect in Nora

Paste the token, optionally set a default model ID (e.g. `meta-llama/Llama-3-8B-Instruct`), and click **Connect**. Nora calls `GET https://huggingface.co/api/whoami-v2` and reports your username on success.

## MCP server

No official Hugging Face MCP server today. Community servers exist for specific use cases (e.g. local Inference Endpoints).

## Environment variables Nora injects

| Variable   | Source          |
| ---------- | --------------- |
| `HF_TOKEN` | API Token field |