LLM API for coding agents: run every AI coding tool all day for $5.50 a week

Nexum Router is an LLM API with heavy, all day usage for $22 a month. No hourly windows, no weekly caps, and all 18 models from DeepSeek, Qwen and Xiaomi included at no extra cost — inside Claude Code, opencode, OpenClaw, Hermes, or any OpenAI-compatible client you already use. Cancel anytime.

Start for $5.50/week →

Included models

• DeepSeek V4.1 Flash — model ID deepseek-v4.1-flash, for fast reasoning, tool use and image input
• Qwen 3.8 Max — model ID qwen-3.8-max (thinking variant qwen-3.8-max-thinking), for flagship coding, agentic work and image input
• Qwen 3.8 Omni Flash — model ID qwen-3.8-omni-flash (thinking variant qwen-3.8-omni-flash-thinking), for fast multimodal work with image input
• Qwen 3.7 Max — model ID qwen-3.7-max (thinking variant qwen-3.7-max-thinking), for flagship coding and agentic work
• Xiaomi Mimo 2.6 — model ID xiaomi-mimo-2.6, the newest Xiaomi model with reasoning and tool use
• Xiaomi Mimo 2.5 — model ID xiaomi-mimo-2.5, for fast iteration and drafting

Plus the wider Qwen 3.5–3.8 family — 18 models in total, all on the same plan.

How it works

1. Create an account, subscribe, and create an API key in the dashboard.
2. Point your tool at the router: OpenAI-compatible clients use /router/v1, Claude Code uses /router/claude.
3. Work without hourly limits or weekly caps.

OpenAI-compatible API setup

Nexum Router speaks the OpenAI API, so the official SDKs, curl, opencode, OpenClaw, Hermes, Cline and similar tools work by changing two settings: the base URL and the API key. Pick a model by its ID.

from openai import OpenAI

client = OpenAI(base_url="https://dialagram.me/router/v1", api_key="dgr_live_your-key")
client.chat.completions.create(model="qwen-3.8-max", messages=[{"role": "user", "content": "Review this diff"}])

Claude Code API setup

Claude Code uses the Anthropic-compatible endpoint: set ANTHROPIC_BASE_URL to https://dialagram.me/router/claude and ANTHROPIC_AUTH_TOKEN to your key. The Claude Code setup guide covers settings.json and choosing the model. Claude Code is the client; the answers come from the open models Nexum Router serves, not from Anthropic's Claude models.

How Nexum Router compares

• vs Claude Code Max — Anthropic's $200/month plan has 5-hour windows plus weekly caps; Nexum is $5.50/week with neither.
• vs opencode Go — opencode Go is $10/month, capped at 20% of a model's monthly allowance per 5 hours and 50% per week; Nexum has no hourly or weekly windows and works inside the opencode CLI.
• vs Ollama Cloud Pro — $20/month for $60 of token-metered usage credits vs $5.50/week with no per-token meter on Qwen and Xiaomi models.
• vs Command Code — credit-metered plans from $1 to $200/month vs a flat plan with no credit meter.
• vs GitHub Copilot — monthly credit allowances ($15–$200) vs a flat plan with no dollar cap.
• vs Cursor — $8 to $200/month tiers on one shared AI usage pool vs a flat plan for terminal agents.
• All Claude Code alternatives compared — the full 2026 roundup table.

Frequently asked questions

How much does Nexum Router cost?

A flat $5.50 per week — $22 a month, counting four weeks. Every model and every supported tool is included; billing is weekly with no contract and you can cancel anytime from the dashboard.

Is there a dollar cap or usage limit?

There is no dollar cap and no hourly or weekly cap on Qwen and Xiaomi models. DeepSeek V4.1 Flash includes $5 of usage a day, renewed daily. Requests are paced per account so the service stays fast for everyone, and the public status page tracks availability.

How do I use it with Claude Code?

Set ANTHROPIC_BASE_URL to the router's /router/claude endpoint with your Nexum API key and keep your exact CLI workflow. Setup takes about two minutes; see the step-by-step guide.

More answers on the full FAQ page.

About Nexum Router · FAQ · Claude Code setup · DeepSeek in Claude Code · DeepSeek in opencode · Status · Support · About Dialagram