wrapper-scioNos / 8.0.0

Your usual AI tools. More choice of models.

Use Claude Code, Codex, OpenCode, Claude Desktop or Cowork with compatible RouterLab models. wrapper-scioNos prepares the connection and guides you through your first launch.

The wrapper is free software. RouterLab model usage follows your access rights and pricing plan.

ScioNos 8.0.0 terminal console with six tool entries, account access and diagnostics.
Console 8.0.0 with example statuses. Your available tools and access may differ.View full size
Open source · license
MIT
Platforms
Windows · macOS · Linux
Requirement
Node.js 24.x

Your tool, the wrapper and RouterLab: who does what?

Your tool is where you work

Keep your familiar application. The model is the AI that answers your requests inside it.

The wrapper prepares the connection

A local terminal console helps you connect your API key, choose your tool and select compatible models.

RouterLab provides the model access

Your RouterLab account determines the available models, credits and consumption. This access is separate from your ChatGPT or Claude subscription.

Which tool do you want to connect?

Choose your tool. Follow only the steps that apply to it.

Guide for version 8.0.0

Claude Code

Choose a Claude, GPT or open-source strategy and a subagent model from the verified options for your RouterLab access.

Before you start

Node.js 24.x and Claude Code installed. Claude Native requires Claude Code 2.1.293 or later; AWS Claude requires 2.1.284 or later.

You also need a RouterLab API key. Your account determines which models appear; a trial account uses Trial variants where available.

Exact requirements and compatibility
  1. Open the guided console

    Open a terminal on your computer and paste this command. Accept npm’s installation prompt if it appears. No global installation is needed.

    Launch wrapper-scioNos shell

    npx [email protected] --service routerlab --lang en
  2. Connect your RouterLab access

    Open Account & access, choose Store or replace token, then enter your RouterLab API key in the masked prompt. Use your API key, not your account password.

    How to get my API key
  3. Choose your models and start

    From Home, choose Claude Code, a model family and the available sub-agent option. Review the summary and launch. With a trial account, choose Trial.

    Claude Code guide
Other installation methods and the separate llm service

Install the package once, then run wrapper-scionos from your terminal.

Keep the command within reach shell

npm install -g [email protected]
wrapper-scionos --service routerlab --lang en

Each service uses its own API key. RouterLab access, credits and permissions do not transfer to the llm service.

llm service — separate access shell

npx [email protected] --service llm --lang en

Compare the models for your usage

AI costs add up as you combine tools, long conversations and repeated agent tasks. The model you choose, token volume and number of exchanges all affect your bill. For some tasks, a cheaper model may be sufficient.

wrapper-scioNos prepares the RouterLab connection and guides you to compatible models authorized for your account. Compare their rates, try the models available in Trial and choose those that meet your needs.

The wrapper supports six tools: Claude Code, Codex CLI, OpenCode CLI, Claude Desktop, Claude Cowork and Codex App. Available models and features depend on each tool.

How do model costs compare with Opus 5.5?

Example of cumulative usage across requests: 1 million input tokens plus 200,000 output tokens, without cache reads, cache writes or additional reasoning tokens. Amounts are in USD at RouterLab catalog rates, excluding plan credit allocations and bonuses.

Example of cumulative usage across requests: 1 million input tokens plus 200,000 output tokens, without cache reads, cache writes or additional reasoning tokens. Amounts are in USD at RouterLab catalog rates, excluding plan credit allocations and bonuses.
Model / API IDExample cost
Claude Opus 5.5claude-opus-5-5Per 1M: input USD 4.00; output USD 20.00.USD 8.00
AWS Claude Opus 5.5aws-claude-opus-5-5Per 1M: input USD 2.40; output USD 12.00.USD 4.80
DeepSeek V4.1 Flashdeepseek-v4.1-flashPer 1M: input USD 0.15; output USD 0.45.USD 0.24
GLM 5.3 Flashglm-5.3-flashPer 1M: input USD 0.10; output USD 0.35.USD 0.17
MiniMax M3minimax-m3Per 1M: input USD 0.24; output USD 0.96.USD 0.432

View the model catalog

This compares token rates at equal volume, not equivalent task results. Actual costs depend on caching, reasoning, response length and the exchanges needed to complete a task. Savings and identical features are not guaranteed. Any subscription required by the tool itself remains subject to its access conditions.

Before your first task

Choosing a model does not guarantee the same features in every tool. Capabilities depend on the model, route and client. Cowork requires the Claude Desktop profile and its local proxy to be running.

Read the technical reference
Is the wrapper free?

Yes. wrapper-scioNos is free, open-source software under the MIT licence. RouterLab model consumption follows your trial or paid access. Any access or subscription required by the original tool remains separate.

Can I use every model in every tool?

No. The wrapper offers models available and authorised for your account and the selected tool. Features and results can differ between models. If catalogue verification fails, launch is blocked rather than silently switching models.

Will this always cost less?

No. Model prices, tokens, caching, reasoning and the number of exchanges determine the final cost. Compare prices, try a first task and review your RouterLab usage. A lower token rate does not guarantee the same task result.

Can I use my ChatGPT or Claude subscription credits?

RouterLab access and consumption are separate from your subscriptions with the original tool providers. A RouterLab API key is needed for the routerlab service.

How do I return to the original Desktop connection?

Use Desktop applications in the console. For Claude Desktop and Cowork, stopping the proxy leaves the managed profile applied; restore official mode to remove it. Codex App has separate restoration controls. The technical reference explains recovery after an interruption.

No models appear or a launch fails. What should I check?

Check Node.js 24.x, your installed tool version, the selected service and that service’s API key. Open Diagnostics in the console. routerlab and llm use separate access. If PowerShell blocks npx.ps1, use the same command with npx.cmd.

Ready for your first task?

Create your RouterLab access, get your API key, then return to the guide for your tool.

Agent Zero, OpenClaw and Hermes can connect directly to RouterLab without the wrapper. View compatible agents

Audience measurement

Allow Google Analytics and RouterLab to measure public visits and link campaign sources to your signup? This is optional and can be changed at any time. If you decline, only anonymous visit totals are kept.