NexusRouter

Overview

NexusRouter is a self-hostable AI gateway. Tools that talk to LLM APIs point at one local endpoint, and the gateway handles the rest: routing each request to the configured provider and model, falling back to a backup when a provider fails or a limit is hit.

Why it exists

Teams using several AI tools end up spreading API keys, quotas, and provider choices across every application. NexusRouter exists to put that in one place: one endpoint, one set of keys, one surface for watching usage and cost.

What is being built

• A single OpenAI- and Anthropic-compatible endpoint for all connected tools.
• Request routing across providers and models, with automatic fallback.
• Per-key budgets with hard limits, plus rate limiting.
• Content guardrails, including PII recognition and prompt-injection detection.
• Response caching and compression to reduce token usage.
• A dashboard for keys, usage, and spending, with ready-made configuration snippets for common AI coding tools.

Technology

NexusRouter is built with Go and ships as a container image, so it runs anywhere a container runs — a laptop, a home server, or team infrastructure. It is open source under the MIT license.

Status

In development. The repository is public and work continues there; the project site carries current documentation.