{"id":1924,"date":"2025-12-18T15:09:57","date_gmt":"2025-12-18T22:09:57","guid":{"rendered":"http:\/\/10kdev.net\/?p=1924"},"modified":"2025-12-18T15:26:38","modified_gmt":"2025-12-18T22:26:38","slug":"creating-a-cli-language-choices","status":"publish","type":"post","link":"http:\/\/10kdev.net\/?p=1924","title":{"rendered":"Creating a CLI: Language Choices"},"content":{"rendered":"

After a lot of my work with gathering JIRA data, I’ve had to go an use the API from PostMan or execute a script more often than now, as opposed to using AI.\u00a0 Even with Claude, Amazon, Rovo, and Gemini — many times will be given a script to execute because the AI’s cannot use my credentials or cannot do a two-step process like gather info, then iterate through that info–> needed arest + code solution.<\/p>\n

Decided to give a shot at creating a simple JIRA CLI (apart from Atlassian’s, which requires special permissions) based on these APIs and scripts I’ve been creating.<\/p>\n

The first thing I am curious about is what language is best to develop a CLI with.\u00a0 Considerations are easy cross-platform deployability, and less set-up (like needing a special version or a code platform).\u00a0 Also, I intend to vide-code mosst of this and do some conversions so would not like any crazy languages as suggestions.<\/p>\n

I asked AI.<\/p>\n

\n

“Best” Language to Write a CLI?<\/h1>\n

Here is a summary that Gemini in GDocs created from Q, Claude, and Gemini results when asked about the “best” code platform to write a CLI:<\/p>\n

\n
\n
\n
\n
\n

Best Language for CLI Development<\/span><\/strong><\/p>\n

<\/div>\n

The document recommends\u00a0<\/span>Go (Golang)<\/span><\/strong>\u00a0as the top choice for professional, cross-platform CLIs.<\/span><\/p>\n

    \n
  • Go (Recommended):<\/span><\/strong>\u00a0The “industry standard” due to its ability to compile to a\u00a0<\/span>single static binary<\/span><\/strong>, which makes distribution incredibly simple, as the end-user does not need to install a runtime environment (like Python or Node.js).<\/span><\/li>\n
  • Rust:<\/span><\/strong>\u00a0The “performance powerhouse,” ideal for system tools and high-performance applications, also offering excellent distribution via static binaries but with a steeper learning curve.<\/span><\/li>\n
  • Python:<\/span><\/strong>\u00a0The “king of prototyping,” favored for rapid development and automation, despite distribution being more difficult due to the requirement of a Python runtime and managing dependencies.<\/span><\/li>\n<\/ul>\n

    Languages for Specific CLIs<\/span><\/strong><\/p>\n

    <\/div>\n

    The document details the languages used by several major CLI tools, which often reflects a balance between the speed of development and ease of distribution:<\/span><\/p>\n

    \n\n\n\n\n\n\n\n\n\n
    CLI Tool<\/th>\nPrimary Language(s)<\/th>\nKey Distribution Factor<\/th>\nNote on Amazon Q CLI<\/th>\n<\/tr>\n<\/thead>\n
    AWS CLI<\/span><\/strong>\u00a0(Amazon’s primary tool)<\/span><\/td>\nPython<\/span><\/strong><\/td>\nV2 is distributed as a self-contained installer that\u00a0<\/span>bundles its own Python runtime<\/span><\/strong>, giving it a single-file-like feel.<\/span><\/td>\nThe document does not specify\u00a0<\/span>Amazon Q CLI<\/span><\/strong>, but the highly relevant AWS CLI is Python-based.<\/span><\/td>\n<\/tr>\n
    Claude Code (CLI)<\/span><\/strong><\/td>\nTypeScript\/Node.js<\/span><\/strong><\/td>\nDistributed via NPM or self-contained installers that can bundle Node.js.<\/span><\/td>\nThe latest version of Anthropic’s open-source tool is primarily written this way.<\/span><\/td>\n<\/tr>\n
    Gemini CLI<\/span><\/strong><\/td>\nTypeScript\/Node.js<\/span><\/strong><\/td>\nDistributed via NPM, Homebrew, or\u00a0<\/span>npx<\/code>, requiring Node.js 20+ for installation via NPM.<\/span><\/td>\nThis is based on the open-source\u00a0<\/span>google-gemini\/gemini-cli<\/code>\u00a0repository.<\/span><\/td>\n<\/tr>\n
    KIRO CLI<\/span><\/strong><\/td>\nTypeScript\/Node.js<\/span><\/strong><\/td>\nLeverages the NPM ecosystem for development.<\/span><\/td>\nThis is a third-party AI CLI tool.<\/span><\/td>\n<\/tr>\n
    GitHub CLI<\/span><\/strong><\/td>\nGo<\/span><\/strong><\/td>\nFollows the industry standard for easy distribution via a single binary.<\/span><\/td>\nMentioned as an example of a popular modern CLI written in Go.<\/span><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/div>\n<\/blockquote>\n<\/div>\n<\/div>\n
    \n
    I found it curious that Gemini didn’t add TypeScript\/Node.js to the list of best choices, being that Claude, Gemini, and KIRO are all writting in that.<\/div>\n<\/div>\n<\/div>\n
    <\/div>\n
    Here is what Gemini had to say:<\/div>\n
    \n
    \n
    \n
    \n
    \n
    Why isn’t TypeScript\/Node.js a top choice?<\/span><\/strong><\/em><\/div>\n<\/blockquote>\n<\/div>\n<\/div>\n<\/div>\n
    \n
    \n
    \n
    \n
    \n

    TypeScript\/Node.js is not generally considered the\u00a0<\/span>top<\/span><\/strong>\u00a0choice for all CLIs (compared to Go or Rust) primarily due to\u00a0<\/span>distribution and runtime requirements<\/span><\/strong>.<\/span><\/p>\n

    The document highlights the following downsides:<\/span><\/p>\n

      \n
    • Requires a Runtime:<\/span><\/strong>\u00a0An end-user must have the\u00a0<\/span>Node.js runtime<\/span><\/strong>\u00a0installed on their system to run a standard Node.js\/TypeScript CLI, which adds a dependency for the user. Go and Rust, in contrast, compile to a\u00a0<\/span>single static binary<\/span><\/strong>\u00a0that runs directly without needing an interpreter or runtime.<\/span><\/li>\n
    • Perceived “Heaviness”:<\/span><\/strong>\u00a0The necessary Node.js dependencies, often contained in a large\u00a0<\/span>node_modules<\/code>\u00a0folder, can make the tool feel “heavy” compared to a self-contained Go or Rust binary.<\/span><\/li>\n
    • Larger Memory Footprint:<\/span><\/strong>\u00a0The Node.js runtime environment generally has a larger memory footprint than a native executable.<\/span><\/li>\n<\/ul>\n

      Despite these drawbacks, the language is still an\u00a0<\/span>excellent choice<\/span><\/strong>\u00a0for specific use cases, such as:<\/span><\/p>\n

        \n
      • Tools for the web development ecosystem.<\/span><\/li>\n
      • Projects that need to leverage the massive\u00a0<\/span>NPM ecosystem<\/span><\/strong>\u00a0and existing JavaScript\/TypeScript tooling.<\/span><\/li>\n
      • Tools where the development team is already familiar with JavaScript\/TypeScript.<\/span><\/li>\n<\/ul>\n

        In modern distributions (like AWS CLI v2 or the Claude Code CLI), this downside is sometimes mitigated by shipping the CLI with an\u00a0<\/span>embedded runtime<\/span><\/strong> so the end-user doesn’t have to install Node.js separately.<\/span><\/p>\n<\/div>\n<\/blockquote>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n

        \n
        \n
        \n

        DevOps vs Developer CLI:<\/h1>\n

         <\/p>\n

        Language Platforms Differ<\/h1>\n
        Finally, I was seeing a bit of a pattern – Go is used (as mentioned by one AI) for Docker, Kubernetes, and Hugo.\u00a0 These are explicitely DevOps-oriented platforms, whereas the coding CLI’s would seem to be mostly used by developers. I asked Gemini about it:<\/span><\/div>\n
        \u00a0<\/span><\/div>\n
        \n
        \n
        \n
        \n
        \n

        So do devops CLIs tend to be written in GO, and application developer CLIs tend to be not written with Go?<\/strong><\/em><\/p>\n<\/blockquote>\n<\/div>\n<\/div>\n<\/div>\n

        \n
        \n
        \n
        \n
        \n
        \n
        \n
        \n

        That is a very accurate observation of the current landscape. There is a strong “language divide” between tools built to manage infrastructure (DevOps) and tools built for application developers.<\/p>\n

        The primary reason for this split isn’t just aesthetic; it’s about\u00a0distribution constraints<\/b>\u00a0versus\u00a0ecosystem integration<\/b>.<\/p>\n

        1. DevOps CLIs: The Reign of Go<\/h2>\n

        If you look at the heavy hitters in DevOps\u2014<\/span>Docker, Kubernetes, Terraform, Helm, and the GitHub CLI<\/span><\/b>\u2014they are almost all written in Go.<\/span><\/p>\n