9router

9router is an open-source, self-hosted AI API routing proxy used by developers to reduce LLM costs and manage multi-provider access from a single endpoint. It sits between AI coding tools and 40+ upstream AI providers — Anthropic, OpenAI, Google Gemini, GitHub Copilot, Kiro AI, Vertex AI, and more — exposing a single OpenAI-compatible endpoint so tools like Claude Code, Cursor, Cline, Copilot, and Codex connect without any reconfiguration. With 7,800+ GitHub stars, it has grown into a widely-adopted tool for developers who want to control their AI API spend without changing their development workflow.

Common Use Cases

Individual developers deploy 9router to route all their AI coding tool traffic through a single proxy, automatically falling back from paid Anthropic subscriptions to cheaper or free provider alternatives when monthly limits are reached — eliminating manual provider switching. Teams use multi-account round-robin to distribute requests across several GitHub Copilot or Anthropic accounts, effectively multiplying their available quota. Developers with large codebases use the RTK Token Saver to compress git diffs and directory listings before they reach the model, significantly reducing per-request token consumption and cost. Budget-conscious users enable Caveman Mode to force shorter model responses on tasks where conciseness is sufficient, cutting output token billing. Security-conscious teams self-host 9router to keep all LLM provider API keys on infrastructure they control rather than trusting third-party proxy services.

Key Features

• RTK Token Saver: Compresses tool call outputs like git diffs, grep results, and directory trees before sending to LLMs, reducing input tokens by 20–40% per request.
• Caveman Mode: Injects terse response instructions into outgoing prompts to reduce model output length by up to 65%, cutting output token costs on applicable tasks.
• Multi-provider fallback: Cascades requests through subscription, cheap, and free-tier providers automatically when quotas run out, with no manual intervention required.
• Round-robin scheduling: Distributes load across multiple accounts per provider to maximise quota usage before monthly resets.
• OpenAI-compatible API: Exposes a /v1 endpoint that any OpenAI-compatible tool can target — no client-side configuration changes needed.
• 40+ provider support: Connects to Anthropic, OpenAI, Google Gemini, GitHub Copilot, Kiro AI, Vertex AI, and dozens more from a single proxy.
• Web dashboard: Browser-based interface at /dashboard for managing provider accounts, routing rules, model combinations, and API keys.
• Usage tracking: Monitors quota consumption across all providers so you can see how much of each subscription has been used before reset.

Why deploy 9router on Hostinger VPS

Self-hosting 9router on a Hostinger VPS keeps all your LLM provider API keys and routing configuration within your own infrastructure — your credentials never pass through a third-party proxy service. You get a persistent proxy URL for all your AI tools, stable enough to set once and leave running, while the routing logic and provider accounts are managed entirely through your own web dashboard. The application runs as a single lightweight container with no external database, making it one of the simplest deployments in the catalogue. All routing rules, usage history, and API keys persist in a named volume, and the service restarts automatically if the VPS reboots.