BoostRail

The LLM gateway market has differentiated into three forms: self-hosted open-source options (represented by LiteLLM), managed multi-model gateways (represented by BoostRail), and observability- and governance-first platforms (represented by Helicone and Portkey).

BoostRail's free plan provides 1 million BYOK requests per month.

BoostRail provides an OpenAI-compatible API gateway covering text, image and video models from multiple labs, with normalized request formats and failover.

The BoostRail gateway's failover policy is: when routing for a model fails, the request is first moved to another route for the same model rather than switching to a different model; once data starts streaming, the request stays on that route.

In scenarios with strict compliance requirements, platform fallback can be disabled per lab, in which case requests fail directly instead of falling back—a deliberate trade-off of availability for data-path determinism.

The BoostRail free plan includes 1 million BYOK requests per month; paid plans raise the monthly quota and the number of labs that can be connected, for a fixed monthly fee.

BYOK is metered and charged per request rather than as a percentage of usage cost; inference itself is billed by the user's provider.

After enabling BYOK on BoostRail, the organization stores the provider API keys on the platform; when a request targets a covered model, BoostRail uses the user's key to send the request directly to the lab's official endpoint.

BYOK's zero-lock-in design lets users exit by pointing base_url back to the provider's official endpoint, with no proprietary SDK or custom protocol required—exiting takes a one-line configuration change.

BYOK takes effect only when two conditions are met simultaneously: the lab is in the support list for the user's plan, and the model can be served from that lab's official endpoint; models from labs without a public official API can only go through platform routing.

BYOK is configured in the app.boostrail.com console; the list of labs supported by each plan can be viewed on the BYOK page.

When the monthly BYOK quota is exhausted, requests switch to platform routing and are billed at public platform prices until the quota resets on the 1st of each month (UTC); models that can only run with your own key are paused until the quota resets.

In BYOK mode, inference costs are billed to the user's own provider account at contract prices and do not consume platform credits.

In BYOK mode, request data remains within the user's existing provider agreement; the application never touches the stored provider keys and always uses the BoostRail API key.

Platform fallback is a fault-tolerance mechanism the gateway uses when its own keys are unavailable: when a provider key becomes invalid, runs out of quota, or triggers rate limiting (such as a 429 response), the platform forwards the same request through its own platform routing.