Enterprise Billing & FinOps¶

Editions: Cloud, Enterprise. Nothing on this page ships in OSS unless it says so.

Preloop Cloud and Preloop Enterprise extend the OSS Cost views with a billing plugin that adds negotiated model pricing, provider billing reconciliation, metered hosted-model analysis, subscriptions, and notifications on budget limits.

OSS owns the ApiUsage ledger, cost summaries, budgets and their enforcement. The billing plugin is gated in the shared console by feature flags.

Feature flags¶

When the billing plugin is enabled, the console exposes additional panels when these flags are true:

Flag	Enables
`billing`	Plan and subscription UI (Settings > Plan) and the Cloud and Enterprise Cost panels
`model_price_overrides`	Account-scoped negotiated pricing for gateway cost estimates
`session_optimization`	Always `true`: session optimization ships in OSS. Cloud meters hosted-model analysis at request time (HTTP 402 via the authorizer hook), never by hiding the UI.

OSS vs Cloud and Enterprise¶

Capability	OSS	Cloud and Enterprise
Spend summaries and trends	`GET /api/v1/cost/summary`	Same
Budgets on the account, flows, API keys and managed agents	`GET/POST/PUT/DELETE /api/v1/budget/policies`	Same
Budgets on users and teams, soft and hard limit notifications	No (HTTP 402)	Yes
Repricing historical usage	`POST /api/v1/cost/reprice`	Same, plus pricing overrides
Model price overrides	No	`/api/v1/billing/cost/pricing-overrides`
Provider billing reconciliation	No	`/api/v1/billing/provider-billing/...`
Session optimization	Yes: analyze / apply / replay at `/api/v1/billing/cost/runtime-sessions/{id}/...` (details)	Same, plus metered hosted-model analysis
Subscriptions and Stripe	No	`/api/v1/billing/subscription`, checkout, portal

Analysis jobs route their own model calls through the Preloop gateway, so that usage is recorded and respects account budgets.

Budget policies¶

Budget policies attach spend limits to a subject: the account, a flow, an API key or a managed agent in every edition, plus users and teams on Cloud and Enterprise. The gateway checks them before dispatch; a priced request that would cross a hard limit gets HTTP 403.

Configure them from an agent's page in the console, or via:

curl -H "Authorization: Bearer $TOKEN" \
  https://preloop.example.com/api/v1/budget/policies

With the billing plugin, crossing a soft or hard limit also sends a notification, and hosted models are priced from their tariff.

Model price overrides¶

Negotiated or self-hosted contract pricing can diverge from public provider defaults. Overrides store per-account rates (per token type, optional request fees, currency, effective dates) so ApiUsage estimated cost reflects your contract rather than list price.

Non-USD contracts: set the override currency and an fx_rate_to_usd conversion rate. Preloop keeps all stored costs in USD (so budgets and cross-account analytics stay comparable) while preserving the original currency and unconverted prices for display and audit. Editing an override can be applied retroactively via usage repricing.

Manage overrides at /api/v1/billing/cost/pricing-overrides or from the Cost view when model_price_overrides is enabled.

Provider billing reconciliation¶

Estimated ApiUsage cost is only ever an estimate. Provider billing reconciliation closes the loop by pulling each provider's own billing/usage actuals and comparing them against Preloop's estimates so you can see and explain estimate-vs-actual drift.

A per-account billing connection links to a provider's billing/usage API; the admin credential is stored via the Secret Service, never in plaintext.
Fetched actuals are persisted as idempotent snapshots so re-fetching the same period does not double-count.
Snapshots are reconciled against estimated spend for the same window.

The shared ProviderBillingConnection/ProviderBillingSnapshot tables live in the open-source models package (one alembic tree), while the fetchers and API endpoints ship in the Enterprise billing plugin.

Session optimization¶

The full optimization loop (analyze, one-click apply, replay verification) ships in OSS and is documented at Session Optimization. What Cloud/Enterprise add on top:

Metered hosted-model analysis: LLM-assisted analysis on operator-hosted built-in models is compute the operator pays for, gated at request time via the optimization_gating authorizer hook (denials return HTTP 402). Deterministic and BYOK analysis are never gated.

Subscriptions and usage metering¶

Preloop Cloud and Enterprise self-hosts can enable Stripe-backed plans. The billing plugin tracks usage against plan limits and exposes checkout and customer-portal flows.

This is separate from gateway spend analytics: subscriptions gate product features while ApiUsage records model consumption.

Contact sales@preloop.ai for Preloop Enterprise licensing.