AI 程式工具導覽

Kiro is an agentic AI IDE from AWS that turns prompts into specs, designs, tasks, code, tests, and pull requests. It is built for teams that want AI coding with more structure than vibe coding, especially on larger or production-oriented codebases.

Zed AI 是一款快速、開源的 AI 程式碼編輯器，其核心圍繞著原生速度的編輯、代理工作流程、自備金鑰（BYOK）模型存取，以及對 AI 生成變更的即時審查。

Lovable 是一款基於瀏覽器的 AI 應用程式建構工具，可以把自然語言 prompts 轉成真實、可編輯的 Web 應用程式。它特別適合 MVP、SaaS 原型、儀表板、內部工具，以及希望能交接到 GitHub，而不是只得到純 no-code 成果的產品團隊。

AnyGen](https://www.anygen.io/%22,%22summary%22:%22AnyGen) is an AI workspace and agent-friendly artifact generator for turning prompts, voice notes, files, and data into polished work outputs. For developer-tool audiences, its main value is the CLI and skill layer that lets coding agents create slides, docs, diagrams, reports, websites, and data artifacts.

ZCode is an agentic development environment from Z.ai/Zhipu AI for long-running coding tasks, multi-step code changes, review, and remote follow-up. It combines a standalone desktop workspace with GLM-optimized agents, BYOK model access, MCP, skills, sub-agents, and mobile/chat control.

Void is an open-source, VS Code-based AI code editor positioned as a privacy-focused Cursor alternative. Its strongest appeal is direct provider connection, local model flexibility, and inspectable source code, but the original repository is now deprecated and archived.

TRAE 包括面向程式碼儲存庫、編輯器與開發任務的 TraeCode，以及面向任務與交付成果的 AI 工作助手 TraeWork。本頁整理產品差異、安裝評估、程式碼審查、方案用量與實作教學。

Windsurf 是一款已轉入 Devin Desktop 的 AI-native code editor，結合了類似 VS Code 的 IDE 與 local/cloud coding agents。它最適合被理解為 agent-first 開發環境，面向希望在編輯器內，而不只是 chat 中使用 AI 協助的開發者。

Google Antigravity 是 Google 推出的 agent-first 開發平台，用於在 IDE、終端機、瀏覽器、桌面 command center、CLI 與 SDK 之間協調 coding agents。它適合希望 agents 透過可審查 artifacts 來規劃、執行、驗證與回報工作，而不只是回覆簡單聊天訊息的開發者。

Theia IDE is an open-source, AI-native development environment for desktop and cloud. It is best understood as both a usable IDE and a reference product built on the highly customizable Eclipse Theia Platform.

Manus is an autonomous AI agent that can research, build, analyze, design, and deploy work products from natural-language instructions. For developers, its strongest fit is prompt-to-app creation, code handoff, browser automation, and workflow orchestration rather than traditional IDE editing.

Orca is a local-first Agent Development Environment for running multiple AI coding agents in isolated git worktrees. It is built for developers who want to coordinate Claude Code, Codex, OpenCode, Cursor CLI, and other agent CLIs from one reviewable workspace.

PyCharm is JetBrains' dedicated Python IDE, combining deep code intelligence, debugging, web and data tooling, and an increasingly open AI layer. It is designed for developers who value project-aware refactoring and reproducible run workflows more than a lightweight, chat-first editor.

GitBot is a local, open-source orchestration hub that turns Claude Code, Codex, or OpenCode into reusable bots with persistent threads and per-bot permissions. It adds a browser workspace and repeatable agent roles without replacing the coding agents you already use.

Oh My Pi (omp) is an open-source, terminal-first coding agent built from Pi and expanded into a batteries-included engineering harness with LSP, debugging, subagents, browser automation, model routing, and deep extensibility.

Cue is Manus's standalone personal-agent app, designed around agents that can hold their own digital identity and carry work across communication, payments, and a dedicated computer. It is a workflow agent rather than an AI IDE or code editor.

Floot MCP turns Claude, ChatGPT, Codex, Cursor, and other MCP clients into full-stack app builders connected to Floot's managed runtime. The assistant supplies the model and reasoning while Floot handles projects, backend services, previews, and publishing.

Cortex AI IDE is a local-first, BYOK agentic code editor that can plan, edit, run, and verify changes across a project. It is designed for developers who want a standalone AI IDE while keeping provider choice and API-key control.

CodexBar is a free, open-source usage monitor for AI coding tools that puts quota windows, reset times, balances, spend, and provider status into a macOS menu bar, Linux desktop app, and CLI. It is designed for developers juggling multiple coding-agent subscriptions rather than for writing code itself.

Nous Research is an open-source AI research lab and developer ecosystem behind Hermes models, Hermes Agent, Nous Portal, and reinforcement-learning infrastructure. For developers, its most relevant products are the open-source Hermes Agent and the multi-model Nous Portal.

A Pi Coding Agent extension that uses TypeSafe Jev as a semantic decision layer for risky tool calls, command-output interpretation, and typed machine-readable judgments.

Codédex is a gamified, browser-based coding learning platform that combines interactive programming lessons, hands-on projects, an in-browser editor, and the Lumi AI coding companion. It is designed primarily for people learning to code rather than developers looking for a production AI IDE.

Follow and control coding-agent sessions on your own computer from an iPhone or iPad. Tactic Remote lets you inspect output, review files, and respond when an agent needs input.

Harden AIF checks supported coding-agent tool calls before execution. It applies local policy decisions such as allowing, blocking, redacting, or asking for confirmation.