Bring your own open model.

You keep the model. You keep the bill.

DreamerOS does not host a model, and it does not pick one for you. You paste your own Hugging Face token once. Every call after that runs on your token, so Hugging Face bills you the way it already does. We do not resell inference and we take no cut of it.

direction  DreamerOS gateway -> router.huggingface.co/v1
auth       your HF token, stored server-side, sent only to Hugging Face
scope      Inference API + Hub read. No write back to the Hub
billing    Hugging Face invoices you. No per-call charge from us
tier       Light (Free)

The connector is live today at app.dreameros.app/connect/huggingface.

Why put anything between you and your model?

A fair question, and the one worth answering first. If you chose open source to keep control, one more service in the path looks like the thing you were avoiding.

So, plainly. DreamerOS does not replace your model, resell it, or choose it. It is the same call you would make yourself, with three things attached that you would otherwise build: what you have already told it, a check that the answer matches what you asked for, and a receipt you can keep. There is nothing to lock you in. The model, the token and the account stay yours, and calling Hugging Face without us works the way it always did.

model      named by you. The gateway never substitutes it
provider   named by you. Hugging Face is one of six here
key        yours, stored server-side, used once, never returned

What runs around the call

Six things happen on the way through, and none of them change which model answers.

Your rough prompt is restructured into a clearer one. Your own fingerprint is attached to the call. Anything you have saved before is pulled in as context. The answer is checked against the intent you actually stated. The output is checked before it reaches you. Then you get a signed receipt for the call.

1 restructure the prompt    4 verify against stated intent
2 apply your fingerprint    5 check the output
3 retrieve your memory      6 sign the receipt

same six on every provider you bring:
openai . anthropic . perplexity . grok . gemini . huggingface

A receipt id can be checked at the receipt page, no account needed.

Set it up in three steps

1. Make a read token at huggingface.co/settings/tokens.

2. Open app.dreameros.app/connect and pick Hugging Face.

3. Paste the token. That is the whole setup.

Hugging Face sits at the Light tier: Free. The only invoice on this path is the one Hugging Face already sends you. Read how connections work before you paste anything.