Private AI for ai builders

One API, every model, published rates

Swap models without touching your code, read every rate before you commit, and meter it against prepaid credits.

What changes

Change models without changing code

OpenAI-compatible and Anthropic-compatible endpoints over the whole live catalogue. Point at a different model id and keep your client library.

Every rate published

Per-token rates for the gateway are public, model by model. You can model your unit economics from a web page instead of a sales email.

Prepaid credits and per-key limits

Meter against a balance you topped up, with limits per key, so a runaway loop in staging cannot drain a month of budget.

What a Tuesday looks like

How Borg serves it

Platform API
The gateway, directly. OpenAI-compatible and Anthropic-compatible, plus image, video and voice endpoints.
The model directory
Every live model has a page with its context window, its modalities, its lane ids and a cited vendor price where one exists.
Prepaid metering
Credit top-ups, per-key limits and a usage endpoint you can read from your own billing code.
Published speed telemetry
Median time to first token and whole-turn throughput per model, from our own request logs, so you can pick on latency rather than on a vendor's claim.

What we can prove, and where it stops

The same two columns as the hub, narrowed to this one business.

Running today

  • The metered per-token business is live and taking real money, with every rate published.
  • Compatible endpoints and a published catalogue of live model ids.
  • Speed figures measured on the same gateway you would be calling.

Not there yet

  • No uptime guarantee and no public status page.
  • Self-serve card signup for the metered product is not live yet.
  • Known routing defects are open and being worked on. Ask for the current list before you build a dependency on a specific lane.

Next step

Start on your own, or ask us the questions this page did not answer.