Claude, withoutthe waiting.
A drop-in replacement for the Anthropic API. Same endpoints, same request shape, same streaming. Point your existing tools at it and keep working.
One command sets up Claude Code, Cursor, Cline, Windsurf and Roo-Code.
POST /v1/messagesclaude-opus-5models
Whole-repo work, long refactors, anything you'd open Claude Code for.
- context
- 1,000,000
High-stakes one-offs where you want it to sit and think.
- context
- 1,000,000
The everyday middle — reviews, edits, tests, glue code.
- context
- 1,000,000
Classification, extraction, and anything in a tight loop.
- context
- 200,000
Every figure here is what GET /v1/models returns — the same public endpoint your client can read. All four take images and extended thinking. Older Claude model names resolve to the nearest tier, so an existing integration keeps working without an edit.
no lock-in
Easy to start. Easier to leave.
There is no SDK of ours to adopt, no message format to convert to, and nothing in your project that only makes sense here. One base URL brought you in. Change it back and your integration still runs — it was never anything but the Anthropic API you already write against.
The only thing keeping you here should be that it works.
don't take our word for it
These three checks are running in your browser right now, against the public endpoints, on a five-second heartbeat. No key, no tokens spent. Open devtools and the requests are there.
- endpoint reachablechecking…
- round tripchecking…
- models availablechecking…
Each bar is one measured reply, scaled against the tallest of them. Relative on purpose — an absolute axis would need a ceiling we would have had to invent.
works with
- Claude Code CLI
- Claude Code desktop
- VS Code extension
- Cursor
- Cline
- Zed
- Claude for Excel
- Claude for Microsoft 365
- Anthropic SDK
- anything speaking /v1/messages
Anything that already talks to the Anthropic API works unchanged — set the base URL and your key, and nothing else in your setup has to move.