Nexum Router

An AI model router for developers and agent users who rely heavily on AI coding tools and regularly run into usage limits. It provides access to 16 models from Qwen, Xiaomi, and Meta through a single API setup, allowing users to switch between different reasoning, coding, and faster iteration models without changing their existing tools.

The service is designed to work with Claude Code, opencode, OpenClaw, Hermes, the OpenAI SDK, and other OpenAI-compatible clients. Instead of configuring separate providers, users can point their workflow at one OpenAI-compatible endpoint and use the included model lineup from a single account.

Key Features

Single API Endpoint

Use one base URL instead of maintaining separate provider configurations for each model family.

• Single endpoint for the available models
• OpenAI-compatible API
• Works with existing compatible applications
• Switch model IDs without changing development tools
• Simplifies multi-model AI coding workflows

16 Models Included

The plan provides access to 16 models across Qwen, Xiaomi, and Meta model families.

The lineup includes options suited to different workloads, including:

• Reasoning-focused tasks
• Coding workflows
• Faster iteration
• Agent-driven development
• General AI coding workloads

All models are included under the same plan rather than being priced individually.

Claude Code & Developer Tool Compatibility

The router is built around tools developers already use for AI-assisted coding and agent workflows.

It works with:

• Claude Code
• opencode
• OpenClaw
• Hermes
• OpenAI SDK
• Other OpenAI-compatible clients

This lets developers keep their existing workflow while changing the model endpoint underneath it.

OpenAI-Compatible API

Applications that already support the OpenAI API format can connect to the router without requiring a completely different integration approach.

This makes it suitable for developers who already have applications, scripts, agents, or coding environments configured around OpenAI-compatible endpoints.

No Hourly or Weekly Usage Caps

The plan is designed for sustained usage rather than short sessions followed by resets.

It includes:

• No hourly cap
• No weekly cap
• No extra per-model pricing
• No add-ons

The service states that its allowance matches the ceiling of a $200/month Claude or ChatGPT-style plan, but across every model in the included lineup rather than being limited to one provider's model stack.

Live Token Usage

The router provides live token totals streamed from the service.

This gives developers a more direct view of actual usage instead of relying only on packaged plan limits or marketing-style usage descriptions.

Account & Usage Management

The service includes tools for managing the account and monitoring usage.

• API key management
• Live service status page
• Usage leaderboard
• Live token totals

These provide visibility into both service availability and how the router is being used.

Weekly Billing

The plan is billed at $5.50 per week.

Users can cancel at any time, with billing ending immediately after cancellation. There are no separate model charges or add-ons described for the included lineup.

Built For Heavy AI Coding Workflows

This router is aimed at developers, AI agent users, and teams who use coding models frequently and care more about sustained access than simply choosing the lowest monthly sticker price.

It is particularly suited to workflows where usage limits interrupt development, agent runs, or repeated coding iterations. Since the models are available through a common endpoint, users can change between model IDs while keeping the surrounding tool configuration intact.

Common Use Cases

AI-assisted software development: Use multiple coding and reasoning models from one API endpoint while working inside existing development tools.

Claude Code workflows: Connect Claude Code to the router and access the included model lineup through the compatible setup.

AI coding agents: Give agent-based development workflows access to different models without maintaining separate provider integrations.

OpenAI-compatible applications: Connect applications that already work with OpenAI-compatible APIs.

Model experimentation: Switch between Qwen, Xiaomi, and Meta models to find the right balance of reasoning depth, coding performance, and iteration speed.

High-volume development: Keep coding workflows running without the usual hourly or weekly usage ceilings described by the service.

Pricing

The plan costs $5.50 per week.

The pricing model includes all 16 available models without additional per-model charges or add-ons. Users can cancel anytime, with billing ending immediately.

Why It Matters

Developers using AI coding tools heavily often run into a practical problem: the model they want to use may be behind a session, hourly, weekly, or provider-specific usage limit.

This router takes a different approach by putting multiple model families behind one OpenAI-compatible endpoint and using a weekly subscription rather than separate model pricing. That means developers can keep their existing tools and switch model IDs as needed, while the service provides live token totals to make usage easier to track.

Keep Your AI Coding Workflow Running

Connect your existing AI coding tools to one OpenAI-compatible endpoint and access 16 Qwen, Xiaomi, and Meta models through a single setup. With weekly billing, no hourly or weekly caps, and model switching without changing your tools, it is built for developers who need sustained AI usage across coding and agent workflows.