Product

Product capabilities

From the first request to production scale, BoostRail covers integration, billing, availability, and team management in one place.

What the platform covers

Integrate once, use the whole catalog.

OpenAI Chat Completions, OpenAI Responses and Anthropic Messages on one base URL. Official SDKs, Claude Code and Codex CLI work by changing the base URL and key.
Streaming uses each format's standard server-sent events, with usage reported at the end of the stream.
Every model in the catalog behind one API key, with no separate account per lab.

One balance for all model spend.

Prepaid USD balance, topped up by card through Stripe, USDT, or Wise bank transfer in USD.
Usage is metered per request, with token counts and the unit price recorded at call time.
An invoice is issued for every top-up, next to the order in the console.
Query and export usage by project, API key or model, and set a low-balance alert in the console.

Your lab accounts, one integration.

Add your own keys for supported labs; those requests run on your own accounts, quotas and contract prices.
Your keys are tried first. Platform routes take over when they fail, are rate-limited or run out, unless you choose to use your keys only.
The free plan includes 1,000,000 BYOK requests per month, with no top-up required.

The same key and balance, for images and video.

Image generation over /v1/images/generations, billed per image.
Video tasks: submit with POST /v1/videos, poll the status, download the result. Billed per second of output (reference images per image).
Failed or cancelled video tasks are not charged; the console shows each task's hold and settlement.
HappyHorse video tasks must request at least 5 seconds.

Built for tools that call models all day.

Claude Code, Codex CLI, Cline, OpenCode, Continue and Aider connect by changing the base URL and key.
Switch models by changing one model id; the rest of the setup stays the same.
Copy-paste configs for each tool are on the Coding page.

Spend your team can see and control.

Organizations, projects and API keys keep workloads apart.
Per-key daily or monthly spend limits and per-minute request limits, set in the console.
Rotate a key, or disable it the moment it leaks.
Request logs and exports support review and reconciliation.

Routes are checked against the official model, and a failed request is retried on another route of the same model.

Each route is compared with the official model at least once a week. A route that keeps failing the check stops serving that model.
When a route returns an error, the request is retried on another route of the same model before any data reaches you. It never switches to a different model.
Requests that fail are not charged.
Requests on your own key can fall back to platform routes of the same model when the key fails, unless you choose to use your keys only.

Create an account, add balance or your own keys, and call any model in the catalog.

Last updated: 2026-10-03