About Hugging Face
Build, train and deploy state of the art models powered by the reference open source in machine learning.
Integration
Hugging Face integrations and automation
Describe what should happen in Hugging Face, and when, in plain words. Whenever builds the workflow and runs it on a schedule, on a webhook, or when another app sends an event.
Connection requirements, provider permissions and optional AI usage apply. Read the runtime limits.
Start for freeNo credit card requiredAI building has weekly usage limits.
Hugging Face on Whenever
How Whenever connects
Through its API, using the credentials and permissions required by the provider.
Build it from your AI client
Connect Claude, Cursor or Codex to the Whenever MCP server and your coding agent can build and deploy the Hugging Face workflow. It runs on Whenever afterwards, on its schedule or webhook, with the chat closed.
Set up the Whenever MCP serverPlan your Hugging Face workflow
Hugging Face shows up in automations in two quite different roles. One is the Hub: a hosting service for models, datasets and Spaces, each a Git-based repository with a model card, tags and download counts. The other is inference: sending text, images or audio to a hosted model and getting a result back, which Hugging Face offers through Inference Providers, a single API in front of partner providers and Hugging Face’s own hf-inference service, plus Inference Endpoints for models you deploy yourself.
Both roles use a user access token sent as a bearer token. Tokens can be read, write, or fine-grained, and fine-grained tokens can be limited to specific resources, such as chosen models or an organisation’s repositories. A workflow that only queries model metadata needs read access; one that pushes files needs write access. Hugging Face recommends fine-grained tokens for production, and some Team and Enterprise organisations require them and reject read or write tokens with a 403.
Inference is billed usage with no markup from Hugging Face. PRO users and Team or Enterprise organisations get monthly credits; free users pay by purchasing credits, and everyone can pay as they go beyond included credits. Requests can also be billed directly by a provider using your own provider key, in which case Hugging Face credits do not apply. Inference Providers covers a selected set of models rather than every Hub repository, so check the model page before writing a brief around it. Generated output can differ between runs, and a model update on the Hub can change behaviour, so note the model id you tested with.
The Hub also sends webhooks. You can watch specific repositories, users or organisations and receive events in scopes such as repo.content for new commits, discussion for new discussions and pull requests, and discussion.comment for comments. Setting a webhook secret makes Hugging Face send it in an X-Webhook-Secret header so the receiver can check the source. That is a better trigger than polling when you want to rerun an evaluation after a model is updated, or alert a team when someone opens a discussion on a dataset they maintain.
Hugging Face workflow examples
Authored starter briefs to configure with your accounts, not preconfigured workflows.
Tag incoming support emails with a classifier
Uses a zero-shot classification model through Inference Providers with fixed candidate labels.
Alert on new discussions in our dataset repo
A Hub webhook trigger instead of checking the repo on a schedule.
Weekly report on trending models for a task
A read-only Hub API query; no inference cost.
Related apps for your workflows
OpenAI
GitHub
Remove.bg
NASA
Perplexity AI
CloudflareHugging Face FAQs
Which Hugging Face token permissions does an automation need?
A read token reads the repositories you can read, a write token can also push to repositories you can write to, and a fine-grained token limits access to specific resources. Hugging Face recommends fine-grained tokens for production.
Is the Hugging Face Inference API free?
No. Free users need to purchase credits. PRO users and Team or Enterprise organisations get monthly credits, and all users can pay as they go beyond them. Hugging Face passes provider prices through without markup.
Can every model on the Hub be called through Inference Providers?
No. Inference Providers serves a selected set of models. For a model it does not serve, deploy it on an Inference Endpoint or host it yourself.
Does Hugging Face support webhooks?
Yes. Hub webhooks watch chosen repositories, or all repositories of chosen users or organisations, and send repo, repo.content, repo.config, discussion and discussion.comment events. An optional secret arrives in the X-Webhook-Secret header.
Will a hosted model give the same answer every time?
Not necessarily. Sampling settings and model updates can change outputs. Record the model id you tested with and keep validation rules outside the model.
Agents should build your automations, not be them
Describe your first workflow in plain words. One stays live for free, no credit card.
No credit card required