Coding
A free coding assistant.
And the API behind it.
Write code, read unfamiliar code, find the bug, explain the stack trace. In the browser when you are thinking, and through an OpenAI-compatible endpoint when you would rather stay in your editor. No per-token bill and no seat price.
Two ways in,
one account behind both.
For the thinking part
Paste a file, describe what it is doing wrong, argue with the answer. Uploads are unlimited and unmetered, so handing over the whole module costs nothing and beats pasting the fragment you guessed was relevant.
For the typing part
Create a key in your account settings and point any OpenAI-compatible client at the endpoint. Nothing to install from us, and no plugin that only works in one editor.
For the part that usually stops you
No per-token pricing, no seat licence and no monthly minimum, so an experiment that turns out badly costs the same as one that works.
It speaks the format
your tools already speak.
The endpoint implements OpenAI-style chat completions, which is the format most editor extensions, agent frameworks and command-line clients already target. In practice that means changing two settings rather than adopting anything: the base URL, and the key you generated. Streaming, system messages, tool calls and structured JSON output all behave the way the clients expect, and the parameters are listed in the API documentation.
The model behind it is Qwen3.8-27B, an open-weights model, which is worth knowing for the same reason the licence on a dependency is worth knowing: you can look up exactly what it is, what it was trained to do, and where it falls down, instead of taking a vendor's word for a black box.
The specifics.
Everything worth knowing before you point something at it.
Questions about the coding API.
Is the coding assistant free?
Yes, in the chat and through the API. There is no per-token charge and no seat price.
Can I use it from my editor instead of the browser?
Yes. The endpoint speaks the OpenAI chat-completions format, so any tool that lets you set a custom base URL and key will talk to it.
Do I need a credit card?
No. You create a key from your account settings and start using it. Nothing asks for payment details at any point.
How is it free?
The model runs on hardware we own, so there is no per-token bill to pass on. Interactive chat is scheduled first, which is what keeps the endpoint open to everybody at no cost.
What languages does it handle?
The usual working set: Python, JavaScript and TypeScript, Go, Rust, Java, C and C++, SQL and shell, along with the config and markup formats around them.
Is my code used for training?
No. What you send is never used to train a model and is never sold. That applies to API traffic exactly as it applies to the chat.
Point something at it.
A key takes a minute to make, and the endpoint looks like the one your tools are already configured for.