Home

Tools your AI assistant can’t do on its own

One free MCP server that works with Claude, ChatGPT, Codex, Cursor, Gemini and more. Copy one line and your assistant gains the tools.

Claude
ChatGPT
Codex
Gemini
Cursor
VS Code
Zed
Cline

Connect in one step

Choose your assistant, copy the line and paste it. There is nothing to install, no account to create and no API key. The remote endpoint gives your assistant four exact tools: generate_qr, count_text, diff_text and check_json. Text you send is processed in memory and is never stored or logged.

Claude Code

Run this in your terminal. It stays available in every project.

Terminal

claude mcp add --transport http --scope user ribox https://ribox.app/api/mcp

Claude.ai, Claude Desktop

Open Settings, Connectors, add a custom connector and paste the address. Works in the browser and in the desktop app. Availability depends on your plan.

URL

https://ribox.app/api/mcp

ChatGPT

Turn on Developer mode under Settings, Connectors, then create a connector with this address. Availability depends on your plan.

URL

https://ribox.app/api/mcp

OpenAI Codex

Run the command, or add the lines to your Codex configuration file.

Terminal

codex mcp add ribox --url https://ribox.app/api/mcp

~/.codex/config.toml

[mcp_servers.ribox]
url = "https://ribox.app/api/mcp"

Gemini CLI

Run the command, or add the entry to your Gemini settings file.

Terminal

gemini mcp add --transport http ribox https://ribox.app/api/mcp

~/.gemini/settings.json

{
  "mcpServers": {
    "ribox": {
      "httpUrl": "https://ribox.app/api/mcp"
    }
  }
}

Cursor

Create this file in your project, or at ~/.cursor/mcp.json to use it everywhere.

.cursor/mcp.json

{
  "mcpServers": {
    "ribox": {
      "url": "https://ribox.app/api/mcp"
    }
  }
}

VS Code (Copilot)

Create this file in your project, or run MCP: Add Server from the command palette and paste the address.

.vscode/mcp.json

{
  "servers": {
    "ribox": {
      "type": "http",
      "url": "https://ribox.app/api/mcp"
    }
  }
}

Zed

Add the entry under context_servers in your settings file.

settings.json

{
  "context_servers": {
    "ribox": {
      "url": "https://ribox.app/api/mcp"
    }
  }
}

Cline

Open the MCP Servers panel, choose Configure and add the entry.

cline_mcp_settings.json

{
  "mcpServers": {
    "ribox": {
      "type": "streamableHttp",
      "url": "https://ribox.app/api/mcp"
    }
  }
}

Any other MCP client

Clients with remote MCP support only need the address. If yours supports only local (stdio) servers, use the bridge entry.

URL

https://ribox.app/api/mcp

mcp.json (stdio bridge)

{
  "mcpServers": {
    "ribox": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://ribox.app/api/mcp"
      ]
    }
  }
}

What your assistant can do

Ribox is a free Model Context Protocol (MCP) server that gives AI assistants exact tools for work they cannot do reliably on their own: changing image files, reading hidden photo metadata, drawing scannable QR codes and counting or comparing text exactly. It does not add tools for things assistants already do well, such as changing letter case.

compress_image: make images smaller, even to a target size such as 200 KB (local).
resize_image: exact sizes, crop, rotate and flip (local).
convert_image: HEIC, JPEG, PNG, WebP, AVIF, BMP and ICO (local).
file_metadata: read or remove EXIF, GPS, XMP and PDF metadata, with verification (local).
generate_qr: real, scannable QR codes as PNG and SVG (local and remote).
read_qr: decode QR codes in images exactly (local).
count_text: exact word, character and sentence counts in any language (local and remote).
diff_text: every difference between two texts, with a unified diff (local and remote).
check_json: strict validation with the exact error line, and lossless formatting (local and remote).

Local server: all nine tools, files stay on your computer

The local server adds the file tools (compress, resize, convert, read or remove photo metadata, read QR codes). It runs on your machine with Node.js 22 or newer, works on files you point it to, makes no network requests and sends no telemetry. Results are written as new files next to the originals unless you ask to overwrite them. The npm package is being prepared and will be announced on this page. Until then, the remote endpoint above covers the file-free tools.

Frequently asked questions

What is MCP?

The Model Context Protocol is an open standard that lets AI assistants call external tools. Adding this server gives an assistant new abilities, such as counting words exactly, while you stay in the conversation.

Is the MCP server free?

Yes. Both the remote endpoint and the local server are free, need no account and no API key.

Which assistants can use it?

Any assistant or editor that supports MCP, including Claude (Claude.ai, Claude Desktop and Claude Code), ChatGPT, OpenAI Codex, Cursor, Gemini CLI and VS Code. The remote endpoint works with every client that supports remote MCP servers over HTTP.

How do I connect it to ChatGPT or Claude.ai?

Add a custom connector in the assistant’s settings and paste https://ribox.app/api/mcp. No sign-in is needed. In ChatGPT custom connectors need Developer mode, and availability depends on your plan.

Where do my files go?

With the local server, nowhere: files are read and written on your own computer. The remote endpoint never receives files, only the short text you ask it to count, compare, validate or encode.

Why is there no tool for changing letter case?

Assistants already do that reliably, so a tool would add nothing. The server only covers work that assistants get wrong or cannot do, like counting words exactly or editing image files.

Do I need to install anything for the remote endpoint?

No. Paste one line in your assistant’s terminal or add the address as a connector. Only the local server needs Node.js, which most developers already have.

Updated: October 14, 2026