Guide updated
Official billing beta for organizations with access already enabled. Real key creation, beta eligibility, signed-in navigation and cost reconciliation await user testing. No new access, credits purchase or role changes for monitoring.
- Required credentials
- Dedicated non-legacy project API key
- Monitoring currency
- USD · costs, not balance
- What is monitored
- Organization costs reported this UTC month
- Fixed GET https://api.together.ai/v1/billing/usage with Bearer authentication; pin organization_id, the current UTC month, granularity=hour and every page. Sum official line_items[].cost USD decimal strings directly: 0.125 is $0.125, not cents divided by 100 or quantity × price. balanceAmount=null. togetherCost.latestWindowEnd is an exclusive UTC endpoint for reported hourly windows, not a final invoice or real-time balance.
Check official billing without sharing a key
- Open your existing organization's Billing and confirm its identity and Credit balance without purchasing or changing auto-recharge to test.
- Use Monthly spend → See current monthly invoice usage, or organization Cost analytics if already permitted. Select the same UTC month and all usage; costs and balance are different.
- Record the amount and time, allowing for completed hours and delay. Without beta access, acceptable key scope or console permissions, stop rather than request a higher role for monitoring.
Connect Together AI · organization costs (beta) to Quota Monitor
Confirm an existing organization and enabled beta
Open Organization Settings below and check the organization that actually pays your Together bill. This API requires billing usage beta to be enabled already. Do not switch organizations, accept invitations, buy credits, request new access or upgrade a plan for monitoring.
Without access, use the manual billing option below. The prepaid platform's existing purchase requirement is not a request from Quota Monitor to pay. This guide does not initiate payments, auto-recharge or a billable model test.
Decide whether the normal key's access is acceptable
Together has no documented billing-only read key. A project key can use resources and spend credits; during beta even an external collaborator's project key can read the whole organization's report. Project scoping does not isolate billing.
An existing project editor can create a key without becoming an admin. Organization Cost analytics in the console is admin-only, unlike this API. If this exposure is unacceptable, stop and check manually; do not invent a read-only policy or bypass approval.
Create a dedicated non-legacy key in an existing project
Open that project's API keys settings, choose Create API Key and use a name such as QuotaMonitorCosts. Optionally select Set an expiration date, choose an appropriate date, then Create. Save the full key privately immediately.
Avoid Deprecated legacy keys; do not regenerate or revoke a business key. A name and expiry do not make the new key read-only. There is no documented IAM JSON import. Keep secrets out of chat, URLs and screenshots.
Privately obtain and verify organization_id
In your own non-shared local HTTP client with cloud sync, history and request logging disabled, create GET https://api.together.ai/v1/whoami. Put your key only in its Authorization: Bearer header, with no body or redirects, never in the URL query.
Copy only organization_id from the successful JSON, not project_id, project_slug, api_key_id or an organization display name. Check the returned organization name against step 1, then close the client. Quota Monitor does not call whoami or collect its identity details. Stop on mismatch or failure; do not switch organizations or copy Cookies.
Add the organization cost connection securely
In your Quota Monitor project, add a connection and select Together AI · organization costs (USD, beta). Enter a recognizable name, the full key in the masked credential input and step 4's organization_id in the required organization field. Currency is USD.
The ID pins the report's organization; it is not project filtering. Each page sends and verifies it. Do not use a URL, project_id or key ID. Credentials are encrypted; the help drawer shares this guide and opening it does not submit a connection or call the provider.
Wait for success and reconcile the same month's costs
After saving, wait for the first sync or use the existing refresh action. Confirm the successful observation time changes. Compare costAmount with the same organization's reported UTC-month costs: 0.125 means $0.125. Preserve fractional and signed adjustments; do not divide by 100.
togetherCost.latestWindowEnd identifies the exclusive UTC endpoint of reported hours. Current-month data may lag by one hour. REST/MCP share this snapshot. A 404, missing field, wrong currency or incomplete pagination fails and preserves the old value instead of becoming zero.
Separate balances, chart filters and internal budgets
In organization Billing, choose See current monthly invoice usage on the Monthly spend card. If you already have organization Cost analytics access, select this UTC month, Cost ($) and All usage. Allow for completed hours and reporting lag, not the default Last 30 Days or a project-filtered chart.
Credit balance, auto-recharge and invoice totals are separate. The current official controls do not provide independent per-key spending limits; a key name or expiry does not create one. Quota Monitor does not read balance or limits; unsupported is not an unconfigured limit. An optional internal budget is a reference difference, not money, a Together cap or guaranteed model access.
Rotate, expire and troubleshoot without expanding access
Create only a replacement non-legacy monitoring key in the same organization, update it through the secure form and revoke only the old monitoring key after a successful sync. Removing its creator from the project does not automatically revoke a key. Leave other application keys alone.
For 401 check the full key, expiry and revocation; for 403 check the key's organization against organization_id; for 404 check existing beta access. Allow 429 backoff. Share codes and times, not raw responses or secrets. Do not debug by paying, elevating roles or guessing balance endpoints; use manual billing while beta changes are investigated.
What Quota Monitor reads from Together AI · organization costs (beta)
- Only existing organizations with billing usage beta already enabled. A 404 is unavailable access, not zero costs. The beta response may change; failures retain the previous successful snapshot and time, without guessed fields or console Cookies.
- Only completed hourly windows in the current UTC month, with up to one hour's lag. Prior-month data may lag 24 hours but is not requested here. Pagination is a moving snapshot, not an atomic invoice. Product, resource, project and key details are discarded; no partial project costs are inferred from attribution.
- No wallet-balance or spending-limit read contract: officialLimit=unavailable/unsupported, not not_configured, unlimited access or zero credits. Optional internal budget differences are reference-only, never funds or a changed provider cap. No runway estimate.
- 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.
Together AI · organization costs (beta) FAQ and troubleshooting
- Does 404 mean no costs this month?
- No. The official billing beta must be enabled per organization; otherwise it returns 404. Keep the old snapshot. Do not try private endpoints, Cookies or a top-up.
- Why can a project key read organization-wide costs?
- That is the documented current beta behavior, including external collaborators' project keys. Admin-only billing access and project isolation cannot be assumed. Use manual checking if unacceptable. Future scope enforcement may require a manual key change, never automatic access expansion.
- Where do I get organization_id?
- Follow step 4 in a private local client using official GET /v1/whoami and copy only organization_id. It is not project_id, a slug or a key ID. Quota Monitor does not call the identity endpoint. For 403, first check the ID belongs to the same key organization.
- Why does a recent request or chart total differ?
- Only completed UTC hours are reported; this month can lag one hour. Last 30 Days, project/key attribution and chart filters differ from organization totals. Compare the same month and all costs with reported-through time. Do not fill gaps with token-price estimates.
- Why are available balance and the provider cap missing?
- cost is reported USD spending, not remaining credits. An unreadable limit is unsupported, not not_configured. An internal budget difference is reference-only, not funds or real-time callable allowance.
- What if the key expires, returns 401 or the beta changes shape?
- Check the dedicated key's expiry, revocation and organization, then rotate safely; allow 429 backoff. Invalid or incomplete pages fail the whole read and preserve the last successful amount/time. Real accounts, console steps and reconciliation remain pending; never share raw responses or secrets.
Official Together AI · organization costs (beta) references
- Official reference · docs.together.ai/reference/billing-usage(opens in a new tab)
- Official reference · docs.together.ai/docs/api-keys-authentication(opens in a new tab)
- Official reference · docs.together.ai/reference/whoami(opens in a new tab)
- Official reference · docs.together.ai/docs/organizations(opens in a new tab)
- Official reference · docs.together.ai/docs/roles-permissions(opens in a new tab)
- Official reference · docs.together.ai/docs/billing-spend-controls(opens in a new tab)
- Official reference · docs.together.ai/docs/billing-cost-analytics(opens in a new tab)
- Official reference · docs.together.ai/docs/billing-credits(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.