welsonjs

WelsonJS Quick Start

한국어 문서

This guide is for WelsonJS 0.2.7.62. It walks you through installing WelsonJS, using the code editor with its AI tools, and running your first script. Developers who want to work from the source code can skip to For developers.

Requirements

1. Install

  1. Download welsonjs_0.2.7.62_setup.exe from the Releases page. It is signed by SignPath Foundation.
  2. Run it and allow it to make changes (it needs administrator rights).
  3. Customer Information (optional): you can leave the fields empty. Your details are sent to Catswords Research only if you fill them in and tick the consent box.
  4. Choose the components (see below), then let the installation finish.

When the files are copied, a PowerShell window runs the post-installation script. It downloads the WelsonJS binaries and the components you chose, which can take a while for the AI models. Do not close this window. When it finishes, the WelsonJS Launcher starts.

WelsonJS is installed in C:\Program Files (x86)\WelsonJS. The downloaded components, including llama.cpp and the AI models, go to %APPDATA%\WelsonJS.

Choose components

The default Full selection includes:

Component What it does
Associate .js files to run with WelsonJS Double-clicking a .js file runs it with WelsonJS instead of Windows Script Host. Clear it if you want to keep the Windows default.
Register the welsonjs:// URI scheme Lets welsonjs:// links start WelsonJS. Any web page can open such a link, so clear it if you do not need it.
WelsonJS Launcher and Windows Service The launcher and code editor, and the WelsonJS Windows service.
Configure WelsonJS MCP server for supported AI CLIs Registers WelsonJS as an MCP server for AI command-line tools that are installed, such as Claude Code.
Python, cURL, websocat Tools that some WelsonJS modules use.
Local AI models (under Llama.cpp) Models that the editor’s AI tools run on your PC. llama.cpp is installed with them.

The Local AI models are about 4.9 GB in total:

Model Download Use
Hy-MT2 1.8B 1.25-bit (Tencent) 462 MB Translation, and the default Local AI model
FrogNano 4B 2609 2.8 GB Coding questions and code writing
BitNet b1.58 2B-4T (Microsoft) 1.2 GB General questions with a 1.58-bit model
Qwen3Guard-Gen-0.6B (Qwen) 484 MB Checks prompts for unsafe content (LLM guard)

Not sure which local models to choose? Enable the ones that match what you need:

To install less, choose Custom and tick only what you need. For example, Hy-MT2 alone is enough for translation. You can run the installer again later to add components.

2. Start the launcher and the editor

The launcher uses its own browser profile, so signing in to web services there does not affect your usual browser. If you upgraded from an earlier version, close the browser windows that the earlier launcher opened first.

3. Write and run a script

The editor starts with an example script. A WelsonJS script exports a main function:

// hello.js
function main() {
    console.log("Hello World");
}

exports.main = main;

Scripts run on the JScript engine of Windows Script Host (based on ES3). Use var and function. ES6+ syntax such as let, const, arrow functions and classes is not available. Modules in the lib folder are loaded with require, for example var HTTP = require("lib/http");.

4. Use the AI tools

The Inference (AI) group in the ribbon has a button that shows the current AI service. Click it to switch to the next service:

Select text in the editor and right-click it to use the following menu items:

Menu What it does
Send Selection Sends the selected text.
Explain Selection Explains the selected code or text.
Find Problems in Selection Looks for bugs in the selected code.
Translate Selection into … Translates into your system language or English. Local AI always translates with Hy-MT2.
Write Code Writes code from your description, in JavaScript that runs on WelsonJS.
Ask… Asks a question of your own.

The answer is inserted into the editor as a comment. Switching the service clears the accumulated conversation.

Local AI

Web chat services

5. Network tools

The Network Tools group looks up information about a host:

6. Change settings

The launcher’s settings are in C:\Program Files (x86)\WelsonJS\bin\x86\WelsonJS.Launcher.exe.config. Edit the file as administrator, then restart the launcher.

Setting Use
IpQueryApiKey Criminal IP API key, for the IP tool
IpQueryApiKey2 AbuseIPDB API key, for the IP tool
DnsServerAddress DNS servers for the DNS tool, separated by commas (default 1.1.1.1,8.8.8.8)
ResourceServerApiKey API key for other software that uses the launcher’s OpenAI-compatible API

The installer replaces this file when you upgrade. Note your changes, and set them again after upgrading.

7. Use the AI from other programs (optional)

The launcher has an OpenAI-compatible API, so programs that support OpenAI can use the Local AI models and the web chat services:

Programs on this PC can use it while the launcher runs. Some programs, such as apps built on a browser engine, need the API key: change ResourceServerApiKey from the public example value to your own key of 16 characters or more. See openai-compatible-api.md for details.

8. Troubleshooting


For developers

You can use WelsonJS from the source code without the installer. Basic use needs no additional software, because scripts run on the JScript engine of Windows Script Host, which Windows includes. WelsonJS runs on Windows XP SP3 or later (verified up to Windows 11 24H2).

Get the source

git clone https://github.com/gnh1201/welsonjs.git
cd welsonjs

You can also download and start it with one PowerShell line. This downloads the master branch into %TEMP% and runs bootstrap.bat:

irm https://catswords.blob.core.windows.net/welsonjs/bootstrap.ps1 | iex

(Optional) Initial setup

bootstrap.bat

bootstrap.bat does two things:

It changes the registry and COM registrations, so it may need administrator rights. Skip it if you use only pure JScript modules.

Run a script

cscript app.js helloworld

app.js loads the module named by the first argument (.js is added when the extension is omitted) and calls its main(args). Extra arguments are passed as an array, and a number returned by main becomes the exit code.

A module and an entry point look like this:

// lib/sayhello.js
function say() {
    console.log("hello");
}

exports.say = say;
// sayhello.js
var SayHello = require("lib/sayhello");

function main(args) {
    SayHello.say();
}

exports.main = main;
cscript app.js sayhello

More

Next steps