MEGA Hub
New
Docs/Routing & billing

Organization Billing#

Organization billing lets an eligible workspace own inference routing preferences, shared custom keys, usage, compute credit, and a monthly hard limit. It is selected explicitly per request and never inferred from repository ownership.

Eligibility and roles#

Organization inference requires a Team or Enterprise plan.

Action Required organization access
Send a request billed to the organization write or admin
View organization inference usage and preferences admin
Change Provider order, disabled Providers, or spending limit admin
Add, rotate, or remove an organization Provider key admin plus token scope account:keys for API automation

Role and plan eligibility apply to every new organization-billed request. After you remove a member or change a plan, allow the settings change to take effect before retrying a request.

Select the billing owner#

Use the organization handle, not its display name:

HTTP
X-Mega-Bill-To: research-lab

Python InferenceClient accepts bill_to="research-lab". Without an explicit billing owner, MEGA uses the caller's personal account even when the model belongs to an organization.

What changes for the request#

Selecting an organization changes the owner of:

  • Provider order and disabled-Provider policy;
  • custom Provider keys used by auto or byok;
  • Routed Inference compute-credit reservation and settlement;
  • monthly inference hard-limit enforcement;
  • usage, token, Provider, and model summaries.

The caller's PAT still authenticates the request and must include inference:run. An organization does not issue a different Router bearer token.

Configure the organization#

Open Organization Settings → Inference Providers. The overview shows organization-only usage; the settings tab controls the monthly limit, Provider order, availability, and shared keys.

For preferred routing:

JSON
{"model": "mega/gpt-5.4-mini:preferred", "input": "Hello"}

Send that body to /v1/responses with X-Mega-Bill-To: research-lab. The Router uses the organization's first compatible healthy Provider. Disabled Providers remain unavailable even when named explicitly.

Combine billing modes#

Organization ownership and billing mode are independent:

Require the organization's saved Provider key:

HTTP
X-Mega-Inference-Billing: byokX-Mega-Bill-To: research-lab

Use mega/gpt-5.4-mini:groq as the request's model value.

Ignore organization keys and use its MEGA compute credit:

HTTP
X-Mega-Inference-Billing: routedX-Mega-Bill-To: research-lab

BYOK requests record zero MEGA inference cost and are excluded from the organization's routed hard limit. The external Provider bills the organization account associated with the saved key.

Automation guidance#

  • Give CI a PAT owned by a maintained service identity with only inference:run.
  • Grant that identity the minimum organization role needed to run requests.
  • Keep X-Mega-Bill-To explicit in every organization-billed workload.
  • Set a monthly hard limit before enabling unattended routed traffic.
  • Restrict account:keys to the separate workflow that rotates shared Provider credentials.
  • Log Inference-Id, organization handle, model, and task, but never prompts or keys.

See Pricing and Billing for reservation behavior and Security and Privacy for data boundaries.