Install Your First AI Agent (Windows 11, Beginner-Friendly)
An AI agent is software that "understands plain language and operates your computer for you": you give instructions in everyday words, and it plans, executes, and delivers results. For bioinformatics analysis, it installs the environment, runs the scripts, and produces the figures — you just copy and paste prompts.
Which one to pick? One-sentence answer:
- Total beginner + Windows 11 → install a desktop (GUI) agent. Recommended: TRAE or WorkBuddy;
- Comfortable with the command line → install a CLI agent. Recommended: OpenCode (open-source, free, works with any model) or Pi (minimal and extensible).
1. Two types of AI agents at a glance
A. Desktop apps (GUI, beginner-friendly)
| Name | Made by | Official site (verified) | Notes |
|---|---|---|---|
| WorkBuddy | Tencent | https://www.workbuddy.cn | All-in-one AI workspace, Chinese UI, ready out of the box |
| TRAE | ByteDance | https://www.trae.ai (intl) | https://www.trae.cn (China) | AI-native IDE, multilingual UI |
| Qoder | Alibaba | https://qoder.com | Agentic coding platform |
| Cursor | Cursor (US) | https://cursor.com | Well-known international AI IDE, subscription-based |
| Windsurf | Windsurf (US) | https://windsurf.com | Well-known international AI IDE, subscription-based |
B. CLI agents (for users with some command-line experience)
| Name | Made by | Site / GitHub (verified) | Notes |
|---|---|---|---|
| OpenCode | SST (open source) | https://opencode.ai | https://github.com/sst/opencode | Free and open source, runs in the terminal, works with any model API |
| Claude Code | Anthropic | https://claude.com/product/claude-code | https://github.com/anthropics/claude-code | Top tier; requires a subscription or API credits |
| Pi | Mario Zechner (open source) | https://github.com/badlogic/pi-mono | npm @earendil-works/pi-coding-agent | Minimal, extensible coding agent |
| Kimi Code | Moonshot AI | https://github.com/MoonshotAI/kimi-code | Official Kimi terminal agent (the older kimi-cli has been superseded by Kimi Code); pairs with Kimi models (API platform: platform.moonshot.cn) |
| Codex CLI | OpenAI | https://github.com/openai/codex | OpenAI's official CLI agent |
| Gemini CLI | https://github.com/google-gemini/gemini-cli | Google's official CLI agent, has a free tier | |
| Aider | Open-source community | https://aider.chat | Veteran open-source pair-programming tool |
Note: CLI agents all require Node.js, are installed via npm in the terminal, and need a manually configured model API key — which is exactly why they assume "a bit of experience".
2. Installing a desktop app: WorkBuddy as an example (recommended for beginners)
- Open the official site (e.g., https://www.workbuddy.cn or https://www.trae.ai) and click "Download" to get the Windows installer;
- Double-click the installer and accept all default options;
- On first launch, sign up / log in as prompted;
- Pick a model from the dropdown next to the input box, and start chatting.
All desktop apps (WorkBuddy, TRAE, Qoder, Cursor, Windsurf) follow the exact same flow: download from the official site → run the installer → log in → pick a model.
✅ Checkpoint: if you can send a message in the input box and get a reply, you're done.
3. Installing a CLI agent: OpenCode as an example (some experience needed)
The universal recipe for CLI agents has only three steps: install Node.js → install globally via npm → configure an API key.
- Install Node.js: download the LTS version from https://nodejs.org and install with default options. Then open PowerShell and type
node -v— if a version number appears, it worked; - Install OpenCode: in PowerShell, run
npm install -g opencode-ai - Configure a model API: run
opencode auth login, pick a provider as prompted, and paste your API key (you can apply for a key on any major model platform — many offer free credits); - In the folder where you want to work, run
opencodeto enter the chat interface.
Other CLI agents are similar: Claude Code is npm install -g @anthropic-ai/claude-code, Gemini CLI is npm install -g @google/gemini-cli, Codex CLI is npm install -g @openai/codex, and Pi is npm install -g @earendil-works/pi-coding-agent. Kimi Code is the exception — no Node.js needed, just a one-line PowerShell script (on Windows, install Git for Windows first):
irm https://code.kimi.com/kimi-code/install.ps1 | iexAfter installation, each has its own login / key-configuration command.
✅ Checkpoint: the terminal shows a chat interface and your messages get replies.
4. About "connecting a model API"
The agent itself is just the "shell" — how smart it is depends on the model behind it. Two ways to connect:
- Built-in (desktop apps): after logging in, you use the built-in models directly — nothing to configure;
- Manual API key (CLI agents): register on a model platform (OpenAI, Anthropic, Google, DeepSeek, Moonshot/Kimi, Zhipu, etc.) → create an API key → paste it into the CLI's configuration. Keys are billed by usage, and many platforms offer free credits — more than enough for everyday analysis.
Note: some model platforms have regional availability restrictions — check the platform's terms of service and supported regions before signing up.
FAQ
Which is the most hassle-free? For total beginners: WorkBuddy or TRAE. With command-line experience: OpenCode.
Can I install several at once? Yes — they don't conflict. Many people keep one desktop app plus one CLI agent.
Is it free? Desktop apps are generally free or have a free tier; CLI agents themselves are free, but model APIs are billed by usage (most platforms have free credits).
What's next after installation? Go back to the analysis tutorial you want to follow (e.g., "Zero to DEG: Bulk RNA-seq Differential Analysis with an AI Agent") and start copy-pasting prompts from Step 1.