Can I connect a personal or Team account?

Not currently. The complete organization report lists Enterprise, Enterprise Plus and Academia Hub. Personal usage-v2 has no complete response contract. Use official billing rather than guessing fields or upgrading.

Why is this not my HF credits balance?

costCents is reported inference spending. Shared credits can fund other Hub products, so subtracting this subset cannot produce a real balance. balanceAmount=null.

Why does a read-only token return 403?

org.billing.read exists in the official permission enum, but the exact minimum combination for this report still needs real-account verification. Check your existing access, organization, plan and token approval. Do not enable writes or a higher role; contact official support or check manually.

Why is this lower than spending just shown in the console?

HF documents up to two hours for requests to appear. Check the UTC month, HF-billed scope and successful timestamp first. Do not call a model or top up to test.

Does an unavailable limit mean there is no cap?

No. unsupported is not not_configured. The integration has no confirmed limit-reading contract. An internal budget is not a provider cap, and its difference is not cash or real-time allowance.

Are members exposed or other sites contacted?

Member, model, provider and token-count fields are not stored or published; only costs are normalized. Fixed huggingface.co organization-report GETs reject cross-host/cross-organization pagination and redirects. Failures preserve old snapshots. Real-account verification remains pending.

Hugging Face inference costs & read-only token guide

Costs only · Not a balance

Guide updated 2026-10-16

Official organization inference report only. Real-account token permissions, signed-in navigation and cost reconciliation await user testing. The public contract lists Enterprise, Enterprise Plus and Academia Hub; personal accounts and Team are not covered.

Required credentials: Dedicated fine-grained hf_ token
Monitoring currency: USD · costs, not balance
What is monitored: Organization inference costs reported this UTC month; Fixed GET https://huggingface.co/api/organizations/{name}/billing/usage/inference with Bearer authentication. Sum daily usage[].costCents for this UTC month and divide exactly by 100 into costAmount; huggingfaceCost.costCents retains the original total. balanceAmount=null. Only the selected organization's HF-billed Inference Providers are included, not personal usage, BYOK, other compute services or prepaid funds.

Check official costs without hosting a token

Choose the target organization, then Settings → Billing; verify its namespace and existing plan.
Compare HF-billed Inference Providers for the same UTC month with up to two hours' lag, not BYOK, Spaces, Jobs, Endpoints or wallet funds.
Record only the time and costs. Personal, Team or insufficient-access accounts use official billing, without paying or expanding privileges for monitoring.

Open official Hugging Face · organization inference costs page

Connect Hugging Face · organization inference costs to Quota Monitor

Confirm an existing eligible organization, not a personal wallet
Open the official organization entry below using the account you actually use for HF billing. Select an existing Enterprise, Enterprise Plus or Academia Hub organization. Personal and Team customers should check billing manually, without buying a plan for this integration.
Record the namespace in huggingface.co/your-org, such as my-organization, not its display name, a personal username, owner/model or a resource-group ID. Confirm your existing billing access; do not change anyone's role.
Open official Hugging Face · organization inference costs page(opens in a new tab)
Open that organization's billing and inference breakdown
From the organization page, open Settings → Billing and its Inference Providers usage breakdown. The organization's Inference Providers settings also show usage by model and provider. Note this UTC month's costs and observation time, without viewing or sharing a secret.
Separate HF-billed requests, Custom Provider Key / BYOK and other Hub compute services. BYOK is charged by the provider directly, while shared compute credits may cover other services. Those totals are not this organization's inference report.
Open official Hugging Face · organization inference costs page(opens in a new tab)
Create a dedicated fine-grained read token
Open Access Tokens below, select New token and Fine-grained, and name it QuotaMonitorReadOnly. In organization permissions, select only the target organization's billing read permission, org.billing.read. Do not enable inference, repositories, member management, payments or org.billing.write.
The official enum contains org.billing.read but does not map this report to a full minimum permission recipe. Personal user.billing.read and OAuth read-billing are not verified substitutes. If your form has no organization billing-read option, contact HF or check manually; do not fall back to a broad read/write token. There is no documented JSON policy import, so no fabricated configuration is supplied.
Open official Hugging Face · organization inference costs page(opens in a new tab)
Save the secret and honor existing approval policies
Save the complete hf_ secret privately at creation, not the token ID, name, header or Cookie. Your organization may require admin approval of fine-grained tokens: check pending/denied in Access Tokens without widening privileges or bypassing approval.
If SSO, MFA or IP policies apply, have the responsible owner verify that server-side reading is allowed. Do not disable policies or add public tunnels. No service account, paid seat, payment method or model test is required.
Open official Hugging Face · organization inference costs page(opens in a new tab)
Add the organization cost connection in Quota Monitor
Open the target project, add a connection and select Hugging Face · organization inference costs (USD). Enter a recognizable name, put the hf_ token in the masked credential field and enter the organization namespace from step 1. Do not enter a URL, personal username, provider model key or resource group.
Currency is fixed to USD. Organization name is the required report scope, not a per-key or project filter. The API encrypts credentials; do not send a secret in chat or MCP arguments. The help drawer uses this same guide and reading it never saves a connection or calls the provider.
Wait for a successful sync and reconcile costCents
After saving, wait for the first sync or use the existing refresh action, then verify the successful observation time changed. This UTC month's daily costCents are summed: 12.5 US cents becomes costAmount=$0.125, never a prepaid balance. huggingfaceCost.costCents preserves the original total; REST and MCP use the same shared snapshot.
Requests may take up to two hours to appear, so recent spending may be absent. No paid model, top-up or permission change is performed. Failures keep the old snapshot and time; missing fields, 403 and incomplete pagination do not become zero spending.
Open official Hugging Face · organization inference costs page(opens in a new tab)
Separate spending, internal budgets and unreadable limits
The primary amount is reported spending, not cash or subscription remaining. Balance is null. The provider's limit is unavailable/unsupported, which does not mean no configured cap, unlimited access or zero credits. Without an internal budget, its difference stays null rather than zero.
If you choose to configure a Quota Monitor internal budget, its remaining amount is only a reference. It does not change HF limits or guarantee model access. Compare the same organization's HF-billed inference in the same UTC month after allowing for reporting lag; do not subtract it from all-Hub invoices or compute credits.
Open official Hugging Face · organization inference costs page(opens in a new tab)
Rotate safely and troubleshoot denied access
Create a replacement monitoring token with the same organization's billing-read scope, replace it through the secure credential update flow, and wait for a successful sync before you revoke only the old monitoring token. Do not touch application keys or other members' tokens.
For 401, check the complete secret, revocation and copying. For 403, check the existing eligible plan, namespace, billing access, organization scope and approval/SSO/IP policy. Allow 429 backoff. Share only codes and times, not tokens, member records or billing screenshots; never raise to write/admin to troubleshoot.
Open official Hugging Face · organization inference costs page(opens in a new tab)

What Quota Monitor reads from Hugging Face · organization inference costs

The public OpenAPI provides a complete inference-cost response only for organizations and marks Enterprise, Enterprise Plus and Academia Hub. Personal usage-v2 has no response/unit contract; no guessed fields or console scraping. Use an existing eligible account, without buying or upgrading a plan.
Requests may take up to two hours to appear. costAmount is reported inference spending, not the final invoice, cash, remaining credits or callable allowance. Compute credits can also pay for Spaces, Jobs and Endpoints; subtracting these inference costs from a plan allowance cannot recover a wallet balance.
No spending-limit read contract is available here: officialLimit is unavailable/unsupported, not not_configured. An optional internal budget is only a reference difference. No key/model/resource-group filtering or runway estimates. Failed pages or invalid daily reports preserve the old successful snapshot instead of publishing a partial cost.
The dashboard shows the latest successful cost snapshot. Refresh cannot remove upstream reporting delays. An old cost and timestamp after a failed query are not a new report.

Hugging Face · organization inference costs FAQ and troubleshooting

Can I connect a personal or Team account?: Not currently. The complete organization report lists Enterprise, Enterprise Plus and Academia Hub. Personal usage-v2 has no complete response contract. Use official billing rather than guessing fields or upgrading.
Why is this not my HF credits balance?: costCents is reported inference spending. Shared credits can fund other Hub products, so subtracting this subset cannot produce a real balance. balanceAmount=null.
Why does a read-only token return 403?: org.billing.read exists in the official permission enum, but the exact minimum combination for this report still needs real-account verification. Check your existing access, organization, plan and token approval. Do not enable writes or a higher role; contact official support or check manually.
Why is this lower than spending just shown in the console?: HF documents up to two hours for requests to appear. Check the UTC month, HF-billed scope and successful timestamp first. Do not call a model or top up to test.
Does an unavailable limit mean there is no cap?: No. unsupported is not not_configured. The integration has no confirmed limit-reading contract. An internal budget is not a provider cap, and its difference is not cash or real-time allowance.
Are members exposed or other sites contacted?: Member, model, provider and token-count fields are not stored or published; only costs are normalized. Fixed huggingface.co organization-report GETs reject cross-host/cross-organization pagination and redirects. Failures preserve old snapshots. Real-account verification remains pending.

Official Hugging Face · organization inference costs references

Prepared by Quota Monitor, not an official provider tutorial. Interfaces and permissions may change; check official pages. Verification is limited to the scope stated here.