Picture by Creator
# Introduction
I’m at the moment attempting to determine which instruments to make use of for my MLOps and vibe coding tasks. There’s a new VS Code extension or command-line interface (CLI) app launching day-after-day, claiming to guide in terminal benchmarks or topping the coding leaderboards. There may be a lot noise within the area that I’m compelled to put in writing this text to share my private experiences with varied Agentic Coding CLI instruments and what I like about them. Please word that these are my private experiences, so they might differ from these of others.
Moreover, the entire CLI instruments talked about under require Node.js, so it’s finest to put in that earlier than testing them. I’ve additionally included the set up instructions that it’s essential kind into your terminal to begin utilizing these instruments.
# 1. Claude Code
Claude Code is a number one instrument for vibe coding and general skilled growth tasks. You possibly can join your Anthropic API keys for token-based utilization or hyperlink your Claude subscription for subscription-based utilization.
I’ve been utilizing Claude Code with the API, and it has labored very well to date. Nevertheless, I just lately found the GLM 4.6 Coding Plan and began utilizing Claude Code with GLM 4.6. This implies you’ll be able to modify your Claude Code to make use of any AI mannequin supplier, together with native AI fashions.
All it’s essential do is run the next command within the terminal to get began:
npm set up -g @anthropic-ai/claude-code
The very best half about Claude Code is that I can simply ask it to make things better or construct parts, and it’ll comply with the directions and supply a brief response. This ends in a low error fee, and it’s fairly proficient at invoking instruments and terminal instructions.
Claude Code is my primary instrument for each day duties, because it additionally comes with a VS Code extension that lets me ask questions on my code and make modifications immediately throughout the IDE. It’s appropriate for each professionals and hobbyists who wish to construct thrilling tasks.
# 2. OpenCode
Opinions on OpenCode are divided; some folks find it irresistible whereas others dislike it. OpenCode is a very open-source various to Claude Code. It means that you can run any mannequin and helps almost all AI mannequin suppliers, providing you with the flexibleness to arrange with out advanced configurations. I exploit it to check new fashions, consider the MCPs, and construct customized brokers.
If you’re a hardcore coder or knowledgeable developer, you’ll admire the intensive customization choices accessible to change and enhance OpenCode at a micro stage. You’ll have management over safety, design, capabilities, and general challenge administration in a managed atmosphere.
To get began, open your terminal and kind the next command:
npm set up -g opencode-ai@newest
You possibly can even discover free fashions supplied by OpenCode and join to numerous fashions via openrouter.ai entry.
I’ve been utilizing OpenCode with the GLM Coding plan and the MiniMax-2 plan. Each have been efficient for my wants, and I exploit OpenCode for constructing the UI of my web site and app.
Observe: If you’re utilizing Home windows, please think about using the Alacritty terminal for one of the best expertise.
# 3. Droid
Droid by Manufacturing unit is an incredible AI coding instrument that’s on the prime of the terminal bench, which means it’s actually good at fixing native issues along with your code. I’ve been utilizing it for debugging and resolving my points, as it may possibly learn the Docker logs, return Docker instructions, and make things better robotically.
What I like essentially the most is that it provides free utilization if you create an account. This implies you get knowledgeable trial plan for a month, which helps you entry the most recent Claude and OpenAI fashions. I’ve been utilizing it day-after-day till my trial expired. I find it irresistible; it’s easy and closely maintained by builders.
To get began, kind the next command within the terminal:
curl -fsSL https://app.manufacturing facility.ai/cli | sh
If you’re searching for accuracy in working instructions, debugging, constructing, and automating your coding setup, I extremely suggest you begin with Droid. The one downside is that it would not work nicely with customized fashions or exterior AI mannequin suppliers. Maybe on the backend, they’ve optimized the mannequin in order that it makes use of Droid to the fullest.
# 4. Codex CLI
I just lately began utilizing the openai/codex CLI, and I did not understand that I may use it with my ChatGPT plan. The ChatGPT plan is way more useful when you might have entry to VS Code extensions, terminal CLI instruments for agentic coding, and cloud-based agentic workflows. For those who want to not pay for the $20 ChatGPT subscription, you’ll be able to nonetheless use it with the OpenAI Developer API.
What’s even higher is that you may modify the configuration file to make use of GLM or Minimax fashions as nicely. It’s totally customizable, however the expertise with exterior APIs can drop considerably, the place it would not perceive sure tags or calls the mistaken instruments. Subsequently, it’s extremely advisable to make use of it with the ChatGPT subscription.
To get began, merely kind the next command into your terminal and comply with the directions:
npm set up -g @openai/codex
I’ve began utilizing it as a result of it’s an virtually free instrument for me to entry the most recent OpenAI Codex and GPT-5 fashions. After Claude Code and Droid, I feel I’ll proceed utilizing it for my tasks. Keep tuned for my subsequent replace on the Agentic CLI instruments quickly.
# 5. Gemini CLI
Google has launched its open-source Agentic CLI, often called google-gemini/gemini-cli, just like OpenAI and Microsoft Copilot. It’s totally customizable, however I encountered many points whereas attempting to set it up and derive any worth from it. For those who by chance use the Gemini 2.5 professional mannequin, it should devour your free plan restrict with only one immediate. Subsequently, I primarily use the Gemini CLI for testing functions solely.
One other challenge I confronted was that, regardless of its customizability, I could not successfully arrange MCP, brokers, or different instruments. I spotted that it already comes with Google service instruments, which inspires the usage of these built-in options as a substitute of counting on third-party instruments.
To begin utilizing the Gemini CLI, kind the next command within the terminal:
npm set up -g @google/gemini-cli
The very best half is that you simply get free entry to Gemini CLI instruments, because the free plan covers your primary wants. It robotically renews, so when you hit the restrict, you’ll be able to start once more the following day. Whereas it’s free, it does include its hassles; for this reason it is listed final. Nevertheless, on account of its reputation and being free to make use of, I made a decision to incorporate it within the listing.
# TL;DR
For individuals who scroll all the way down to learn the abstract:
- Claude Code is one of the best; use it as your major instrument
- OpenCode is wonderful when you like tinkering along with your workflow to maximise productiveness and worth
- Droid is actually good at debugging and automating your developer expertise
- Codex CLI is bettering and now consists of options that help you construct issues each domestically and within the cloud utilizing the ChatGPT plan
- Gemini CLI is widespread on account of its restricted free plan, however I might extremely discourage utilizing it, as you could waste time and acquire little worth from it
There are different AI coding CLI instruments on the market, however they aren’t mature or widespread sufficient for me to incorporate right here. All of the CLI instruments talked about above work flawlessly on Home windows, even with out Home windows Subsystem for Linux (WSL). So, go forward and begin typing the above instructions to expertise the brand new period of agentic coding.
Abid Ali Awan (@1abidaliawan) is an authorized information scientist skilled who loves constructing machine studying fashions. At present, he’s specializing in content material creation and writing technical blogs on machine studying and information science applied sciences. Abid holds a Grasp’s diploma in expertise administration and a bachelor’s diploma in telecommunication engineering. His imaginative and prescient is to construct an AI product utilizing a graph neural community for college kids scuffling with psychological sickness.
