™Bring your own key.
Connect an OpenAI-compatible endpoint with your own credentials. Choose models and manage billing with your provider.
- Keep your provider relationship
- Configure chat and embedding endpoints
- Test compatibility with your workflow
Choose your AI models.
Connect the models your app needs. LibraOS handles agents,
memory, knowledge, and tools around them.
YOUR APPThe experience
LIBRAOS RUNTIMEThe execution
MODEL ENDPOINTThe inference
Choose chat and embedding models for your workload.
Connect your model.
Keep inference local.
Connect a compatible model server on your machine or private network, alongside your LibraOS Runtime.
- Select models for your hardware
- Control model and data placement
- Validate network access and connectors
Use hosted inference.
Discuss hosted model inference for your application. Confirm the available models, API access, pricing, and capacity with our team.
- Match models to your workload
- Review serving and capacity needs
- Discuss latency and cost requirements
Start with the model card.
Explore the model repositories and check the base model, publisher, license, modifications, and deployment requirements before choosing weights for your application.
Model weights: what the labels mean
Fine-tuned models
Additional training adapts a model. Check who performed it, the intended use, and evaluation results.
Quantized models
Reduced precision changes the memory and hardware tradeoffs. Check the quantization method and its author.
Model mirrors
A hosted copy may preserve the original weights unchanged. A mirror is not a new LibraOS fine-tune.
Put your model to work.
Your chosen model powers responses. LibraOS connects those responses to persistent context, knowledge, and actions. You can use the Runtime with your own model provider.