One OpenAI-compatible endpoint for OpenAI, Anthropic, Kimi, GLM, Qwen and other open-source models, routed through the EU wherever possible, and for your own agents.
Switch providers by changing a model name, not your code.
Claude 
OpenAI 
Mistral 
Llama State-of-the-art open-source models are served from European data centers. Proprietary models are routed through the EU wherever the provider allows it.
One key in the dashboard covers every model and every agent you run on zwrm.
Keep the OpenAI SDK you already use. Change the base URL and pick a model name, or let zwrm/auto choose.
Let the router pick an EU model for the job, pin a specific model for a task, or send the prompt to one of your own agents.
One endpoint means one bill, one set of limits, and one place to see what is actually being used.
Chat completions, streaming, and tool calls work with the SDKs and frameworks your team already uses.
Open-source models run in European data centers. Proprietary models are routed through the EU where the provider allows it, and labeled when they are not.
zwrm/auto sends each prompt to a model that fits it: a quick summary to a fast EU model, a refactor to a code model.
Name a model explicitly for tasks where a specific provider or capability is non-negotiable.
Address your own zwrm agents like any other model. A prompt becomes a run with tools, memory, and a workspace.
Usage per model and per key in the dashboard, so the bill is never a surprise.
Use the managed endpoint on European infrastructure or run the zwrm model gateway in your own data center, pointing at the providers you approve.
Discuss your deploymentOpenAI and Anthropic models, plus open-source models such as Kimi, GLM, and Qwen hosted in Europe. The dashboard lists every model with its context window, price, and region.
Yes. Use the official OpenAI SDK or any framework that speaks the chat completions API, change the base URL to zwrm, and use your zwrm API key.
Open-source models run on European infrastructure. Proprietary models are routed through EU endpoints wherever the provider offers them; models that are not EU-hosted are labeled as such so you can exclude them.
It picks a model for each request based on the task, preferring EU-hosted models. If you need a specific model, name it explicitly and the router steps aside.
Every zwrm agent is addressable as a model name. Sending a prompt to it starts a run with that agent's tools, memory, and workspace and returns the result like a normal completion.
Yes. Connect your existing OpenAI or Anthropic account and zwrm routes those models through your subscription while keeping one endpoint for everything.
Create an API key in minutes, or talk with us about routing rules, data residency, and volume pricing.
Hi, I'm Vonk. I answer questions about zwrm and help you plan your agent setup. What do you want to know?