Valefront is an LLM router created by a small team in Da Nang. You point the OpenAI or Anthropic SDK to llm.valefront.com, send "model": "auto", and each request is routed to the model suitable for the task, quality, cost, and latency, with automatic failover when the provider is slow or fails.

The problem we encounter: each feature is hardcoded to a model, paying high prices for prompts that smaller models handle well, and chaos arises when the provider is slow. Furthermore, many teams in Vietnam find it difficult to make payments in USD and need VAT invoices for each expense.

Main features:

• Compatible with OpenAI and Anthropic: Claude Code, Codex, OpenCode, Pi just need to change the base URL
• Choose routing strategy: cheapest, fastest, best, balanced, or set your own rules
• Budget per key, hide personal data (PII), view costs per request
• Pre-load, charge the actual official price from the provider, no monthly fees
• Pay in VND via bank QR, with a VAT invoice

Who should use it: developers and software teams calling AI APIs in products, or using a coding agent daily.

How to get started: register at dash.valefront.com, create an API key, change the base URL in the code or coding agent. Get 1 million tokens for a trial, plus 4 million after the first load.

Difference: not just an API gateway, Valefront chooses the model based on the strategy you set, and payment, invoicing, and support are all tailored for Vietnamese customers.

Launched on November 7, so there are still many rough spots. If you encounter errors, click "Report Error" in the dashboard, or join Valefront’s Discord to provide feedback.