Two pages went live on apispi.com this week, and they both answer the same underlying question we kept hearing: what exactly can I point at this thing, and how? The Model Library answers it for models. The new Compatible Clients directory answers it for tools.
The Model Library, Now With a Page Per Model
The Model Library already listed every provider ApiSpi can route to — Anthropic, Gemini, DeepSeek, Grok, and the rest. What's new is that every individual model now has its own page: pricing, a short description, and a stable URL you can actually link to or bookmark, instead of a name buried in a provider card.
The bigger change is underneath the page, not on it. Pricing used to live in a config file that only shipped with a code deploy. It now lives in a database table our admin team edits directly — which sounds like a small technical detail, but it means the price shown on the pricing page is the exact same number the gateway bills your Token Bank at, always, because they're now reading from the same source instead of two files that could quietly drift apart.
Compatible Clients: A Directory, Not Just a List
The gateway has always been OpenAI-compatible — point anything that lets you set a custom base URL and API key at it, and it works. That's true, but it's not obviously actionable if you don't already know which of your tools has that setting, or what it's called. The new Compatible Clients page turns that into a real directory: terminal agents (Claude Code, Codex CLI, Cline, and others), IDE integrations like Cursor, workflow tools like n8n, and chat platforms — grouped by category, each with a one-line description.
Click through any of them and you land on a dedicated setup guide with the gateway's base URL, a prompt to grab your API key, and — specific to that tool — exactly where that base URL and key go. A handful of the most common tools get precise, verified instructions; for newer or less-documented ones, the guide is upfront about pointing you to that tool's own docs for the exact setting name rather than guessing and risking a broken config.
One Gateway Underneath Both
Neither page introduces a new integration surface. Every model in the library and every client in the directory ultimately talks to the same single endpoint — the LLM Gateway — so nothing here changes how requests are billed, governed, or tracked. It's the same Token Bank, the same governance guardrails on every message, the same usage tracking in your dashboard. What changed is how easy it is to find the model you want and get the tool you already use talking to it.