Guide updated
Based on official documentation; real-account verification is pending. Please test key creation, permissions, organization scope, amounts and these steps. Automated tests are not live reconciliation.
- Required credentials
- Featherless organization Admin API key
- Monitoring currency
- USD · US dollars
- What is monitored
- Available organization Credits · USD
- Fixed GET https://api.featherless.ai/credits/balance. available_nano_usd / 1000000000 supplies available USD, with total and reserved Credits shown separately (scope=organization). The official example available_nano_usd=42488000000 means $42.488.
Check Credits without sharing a key
- Sign in to Featherless and open your account dashboard's Billing page.
- Select the paying organization and compare Developer USD Credits, not a fixed chat-plan allowance or Hugging Face bill.
- Note total, reserved/available amounts where shown, and the time. Allow for settlement lag; no key needs to leave Featherless.
Connect Featherless AI to Quota Monitor
Confirm the organization and billing mode
Open Featherless below and sign in yourself. From your account dashboard, open Billing and confirm the organization that pays for your API use uses Developer/request pricing.
Check: record its USD Credits and the time. Do not switch plans, subscribe or add funds merely to test this monitor.
Confirm existing Admin permission
Ask the organization owner to confirm that the key owner already has the Admin role. The balance endpoint checks manage_billing; Editor, Developer and Viewer keys cannot read it.
Check: your organization approves storing this privileged credential. There is no documented balance-only policy to paste; do not promote members solely for monitoring.
Create a dedicated API key
Open the API Keys page below under the approved account/organization. Click Create New API Key, enter Quota Monitor as the name, then click Generate.
Check: this is a new dedicated key, not a reset of a production key. If the controls differ or access is missing, stop and confirm with your administrator.
Save the secret safely
Copy the complete secret when shown and save it in a password manager; the official setup guide says it is shown only once. Copy only the value, without Bearer, Authorization:, quotes or whitespace.
Check: keep the key out of chat, screenshots, URLs and MCP arguments. No organization ID or custom API host is needed.
Connect inside your project
In Quota Monitor, open the target project and add a provider connection. Select Featherless AI, enter a recognizable name, paste the key into the masked field and save. MCP users use the secure setup link.
Check: USD is fixed and project/key filters are unsupported. Linking to another project does not partition the upstream funds.
Wait for a successful sync
Wait for the first automatic sync or request one manual refresh. Check both the success status and updated timestamp; respect the refresh cooldown.
Check: a retained old amount after failure is not a new result. Null or missing values must not appear as zero.
Reconcile available, total and reserved Credits
Return to the same organization's Billing page. Compare like-for-like at similar times: the example total is $42.5, reserved is $0.012 and available is $42.488.
Check: featherlessBalance.balance is available, not total. Allow for the five-minute settlement window and in-flight requests. Never estimate a correction or subtract reservations twice.
Keep access safe and complete acceptance
Use Open provider to return to Featherless, then your account's Billing page when funds run low. Only the authorized account owner should decide whether to pay.
Check: rotate via a new approved key, verify sync, then revoke the old key. Removing a project link or deleting this connection does not revoke the provider key. Report only redacted errors and timestamps.
What Quota Monitor reads from Featherless AI
- Developer plan/request-pricing organization Credits only, not withdrawable cash, a project wallet or an individual key limit. Several connections can share the same funds; do not sum them.
- Reservations are already deducted from available funds. Do not deduct reserved_nano_usd or usage again. Integer nano-USD fields preserve exact precision; display _usd strings do not replace them.
- Settlement uses five-minute windows and may lag live traffic. A fresh query does not guarantee fully settled costs or model access. Fixed chat plans and dedicated GPU contracts are not prepaid API balances.
- Hugging Face-routed calls billed by Hugging Face are excluded. Use a Featherless key, not an hf_ token. Stored Credits on a chat plan do not represent its remaining allowance.
- REST/MCP featherlessBalance.balance, totalBalance and reservedBalance are exact USD strings; settlementBasis=five_minute_windows. Missing/null or inconsistent amounts fail without replacing the last successful snapshot. Zero and negative available values remain valid.
- The dashboard shows the latest successful snapshot. Manual refresh queries the provider again but cannot eliminate upstream billing delays. An old balance is not a new result after a failed query.
Featherless AI FAQ and troubleshooting
- 401: key rejected?
- Re-copy the dedicated secret from API Keys, check whether it was revoked, and remove header prefixes or whitespace. Do not substitute a Cookie or hf_ token.
- 403 / organization_permission_denied?
- The key must be owned by an organization Admin with manage_billing. Model access alone is insufficient. Confirm the key's organization and existing role with its owner; do not widen permissions or use a session Cookie.
- Available differs from the headline balance?
- The headline may show total Credits. Available excludes reservations for in-flight work. Compare the same organization and time; do not deduct reserved funds again.
- Missing/null amounts or invalid_response?
- All three integer nano-USD amounts and currency=usd are required. Malformed or inconsistent data keeps the previous successful amount and timestamp. Share only the redacted error.
- 404, 429 or timeout?
- Wait before retrying and check the displayed timestamp. The only supported path is /credits/balance, not /v1/credits/balance. Do not try mirror hosts or expose the key in a URL.
- Can this monitor fixed plans or Hugging Face billing?
- No. Chat/dedicated capacity and Hugging Face-routed charges have different billing scopes. Credits can remain stored when switching plans, but are consumed only under request pricing; they are not unused subscription quota.
Official Featherless AI references
- Official reference · featherless.ai/docs/api-reference-credits(opens in a new tab)
- Official reference · featherless.ai/docs/request-pricing-and-credits(opens in a new tab)
- Official reference · featherless.ai/docs/guides-roo-code(opens in a new tab)
- Official reference · featherless.ai/docs/api-overview-and-common-options(opens in a new tab)
- Official reference · featherless.ai/docs/hugging-face(opens in a new tab)
- Official reference · featherless.ai/docs/billing(opens in a new tab)
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.