Home → Blog

Blog

Short, specific, and only when there is something to say.

What gets written here

Things we hit running this platform that are not documented anywhere else, and that cost someone real time or real money before we understood them. Payment mechanics, billing behaviour that surprises people, and the failure modes that look like something they are not.

What does not get written here: model announcements rewritten from a provider's press release, capability comparisons we have no data to support, and posts that exist to carry a keyword. If a page has nothing in it you could not get from the provider's own documentation, it should not be a page.

The three things that surprise people most

If you are new here and only read one thing, make it one of these. Each has cost somebody an afternoon:

1. A wrong base URL produces an error that appears in no log. The 404 is returned at the edge, so the request never reaches the gateway and your usage history stays empty. From the client it looks identical to a bad key or a missing model. The rule is /v1 for OpenAI-lineage clients, no /v1 for Claude Code and the Anthropic SDK — full settings per client.
2. A key belongs to exactly one group. A key made for text models is refused by image models even though the key is perfectly valid. If some models work and one does not, on the same key, that is a group boundary and the fix is a second key, not a permission change.
3. Cache direction moves the bill more than the model tier does. Cache reads cost a fraction of the input rate; cache writes cost more than it. One real session on this platform ran 30 requests for $1.13, and 63% of that was cache writes — context being rebuilt, producing no output at all. The arithmetic is here.

Elsewhere on the site

Setup and reference material lives in Guides rather than here — base URL configuration, what OpenAI-compatible actually covers, Claude pricing and cache economics. Errors and what they mean are under Troubleshooting. Workload-specific cost breakdowns for coding agents, retrieval pipelines and support bots are in Use cases.

Prices are on the price list, regenerated automatically from live billing data rather than maintained by hand, so a published figure cannot drift away from what the system actually charges.

Start using it

No subscription, no monthly minimum, no sales call. Top up with USDT and spend what you use — 1 USDT gives you 2 credits of API balance.

Read the 30-second quickstart

Prices quoted on this page are regenerated automatically from live billing data. Third-party terms are quoted from that party's own published documentation.