- Figma
协作式界面设计，支持组件、变量、原型与 Agent 辅助画布编辑；可通过 MCP 连接设计和代码工作流。
- Figma Make
从提示词和设计参考生成交互应用，可编辑代码、预览与发布；预览可复制为设计图层，席位和额度依方案而异。
- FigJam
Collaborative whiteboard for workshops, diagrams, and AI-assisted brainstorming.
- Framer
可视化网站设计、CMS 与托管发布平台；官方不支持导出整站 HTML 自托管。需要源码交付或独立托管时，应先确认替代方案。
- Sketch
Mac-first UI/UX design and prototyping app with mature component workflows.
- MasterGo
Collaborative UI design and prototyping platform with strong China-market adoption.
- Antigravity
Google 的 Agent 开发环境，连接编辑器、终端和浏览器工作；方案与额度以官网及账户实际可用功能为准。
- ChatGPT
OpenAI’s assistant for writing, research, coding, and multimodal tasks, available on web, mobile, and desktop.
- Claude
Anthropic’s assistant, strong for long documents, structured writing, and careful analysis.
- Gemini
Google’s multimodal model family and Gemini chat experience, with developer access via Gemini API, AI Studio, and cloud products.
- Perplexity
Answer engine that combines live web retrieval with numbered citations—suited to verification-heavy Q&A.
- Grok
xAI assistant with live web search, citations, and tight X integration; free tiers and paid plans differ by limits.
- DeepSeek
Open-weight model family with strong reasoning/coding capabilities and hosted options.
- Doubao
ByteDance consumer AI assistant for Chinese-language chat, writing, and daily tasks; models are also available via Volcano Engine APIs for developers.
- Kimi
Moonshot AI assistant for long-context chat, documents, and multimodal tasks; web/mobile apps plus a separate Open Platform for API access.
- Midjourney
用于概念图、情绪板和视觉风格探索，支持参考图、网页编辑与图片动画。适合先确定视觉方向，再尝试短片运动；交付前检查主体细节与镜头连续性。
- Runway
通过生成工具和对话式 Agent 制作、编辑视频，可在时间线上整理片段。适合分镜和视觉动效探索；超过所选模型时长的内容会分段生成，需检查镜头衔接与主体一致性。
- GPT Image
OpenAI 图像生成与编辑现应参考 GPT Image；DALL·E 2/3 API 已于 2026-05-12 退役，旧链接保留用于兼容。
- Stable Diffusion
Image-model ecosystem with local and hosted workflows; licenses vary by model and use.
- Ideogram
适合含文字的海报、封面和社交图设计，可用 Layerize Text 提取文字图层后修改文案、字体和字号。该功能仍为测试版，弯曲或装饰性文字可能识别不全，交付前需校对。
- Nano Banana
Gemini 原生图像生成与编辑系列，包含 Nano Banana 2、2 Lite 和 Pro；按参考图一致性、编辑需求与成本选择。
- Pika
用于短视频生成和特效探索，可通过 Pikaframes 指定首尾画面，或用 Pikaswaps 替换视频中的对象。适合社交内容与转场试做；生成时长和输出规格取决于所选功能与模型。
- Claude Code
Anthropic 的编码 Agent，可读取项目、修改文件、运行命令并连接开发工具，覆盖终端、IDE、桌面和浏览器。
- Cursor
具备项目上下文的 AI 编辑器，Agent 可修改多个文件和运行工具；以代码差异与真实运行结果验收。
- VS Code
Mainstream code editor and extension platform used as the base environment for many AI coding workflows.
- Codex
OpenAI 的编码 Agent，覆盖桌面、CLI、IDE 和云端工作流，可使用技能与集成完成修改、检查和评审。
- GitHub Copilot
覆盖补全、对话和 Agent 工作流；云端 Agent 可异步处理任务并准备 PR，仍需代码评审与测试。
- Devin Desktop
原 Windsurf，现为 Cognition 的 Devin Desktop：在 IDE 中管理本地和云端编码 Agent、上下文及代码评审。
- Lovable
通过自然语言构建应用，并可通过 GitHub 导出和双向同步项目代码。同步代码后仍需配置部署环境与数据服务；内置安全扫描不能替代完整的应用验收。
- Replit
带 AI Agent 的云端开发与发布平台。发布会创建独立于编辑器的应用快照；修改后需重新发布，持久化数据应使用数据库或存储服务，不能依赖部署文件系统。
- Bolt
在桌面 Chromium 浏览器中生成网站与应用，支持 Node.js 后端和 Expo 移动应用流程。原生项目后端以 JavaScript 为主，不支持直接运行 Python 或 PHP 后端。
- v0
从提示生成界面和全栈应用，默认使用 Next.js，可连接数据库与 API 并发布到 Vercel。按界面、数据、权限逐步实现；预览通过后仍需验证生产配置和真实交互。
- Notion
文档、知识与项目协作空间，支持 AI Agent 工作流；检索和操作范围取决于连接内容、工作区权限与方案。
- Obsidian
Local-first Markdown notes and personal knowledge base with links, graph view, and plugins.
- Gemini Notebook
原 NotebookLM，2026 年 7 月更名为 Gemini Notebook：围绕来源资料进行研究、整理与问答，并与 Gemini 笔记本连接。共享笔记本与聊天上下文有不同规则；引用仍需回查原文。
- Google AI Studio
Build and test Gemini prompts/APIs with tooling for developers.
- Anthropic SDK
Open-source SDKs (MIT) to integrate Claude; API usage is billed per token separately.
- Vercel AI SDK
TypeScript toolkit for building AI apps with unified provider APIs.
- LangChain
Framework for building LLM apps and agents with integrations and tooling.
- Ollama
本地运行与管理模型，也可连接 Ollama 云端模型。适合本地原型和模型试用；需要离线工作时，应确认使用本地模型并关闭云端功能。
- Open WebUI
可自托管的 AI 对话与知识库界面，可连接本地运行环境或外部模型 API。适合统一模型入口；是否完全本地取决于连接方式，修改品牌标识需符合许可证条件。
- Llama
Meta open-weight model family used across local, cloud, and hosted deployments (see llama.com for releases and Llama API).
- Mistral
Open-weight model family and hosted APIs from Mistral AI.
- Qwen
Qwen model family with downloadable weights and hosted APIs; licenses and costs vary by model and provider.
- Gemma
Google open-weight model family for local and hosted workflows under model-specific terms.
- Phi
Microsoft small-language-model family for local and hosted experiments; check each model’s requirements and license.
- Uizard
通过提示词、草图或截图制作可编辑设计稿与原型；不能把整个项目导出为 HTML/JavaScript 应用。
- Lucidspark
Enterprise whiteboard for brainstorming, mapping, and workshop facilitation in cross-functional teams.
- Miro AI
Enterprise whiteboard with AI sidekicks and flows. Common in large-org collaboration.
- Copilot
Microsoft AI assistant across Windows, Microsoft 365 apps, Edge, and web—focused on everyday productivity and work scenarios.
- Adobe Firefly
创意生成与编辑平台，包含 Adobe 与合作方模型；不同模型的训练数据、使用条款和保障条件需分别核对。
- ComfyUI
用节点组织图像、视频等生成工作流，支持多种模型和 API 调用。适合需要复用流程与细调参数的创作；本地运行需匹配硬件，模型与扩展各有使用条件。
- Luma Dream Machine
从文字或图片生成视频，也可用 Ray3 Modify 改造已有片段并加入角色参考。适合探索不同造型和画面风格；检查动作与角色一致性，接入 API 前另行确认所需模型和功能。
- Kling AI
AI video generation for exploring motion and visual sequences.
- Veo
Veo: Google DeepMind generative video (e.g. Veo 3.x) reachable through Gemini apps, Flow, and the Gemini API.
- Seedance
ByteDance Seedance AI video: text/image to video with strong motion and audio sync; distributed via CapCut, Dreamina, and related ByteDance surfaces.
- Descript
Edit video and podcast like editing text. Perfect for creator pipeline and content ops.
- OpusClip
Long video to shorts in one click. Optimized for TikTok, Reels, Shorts.
- Meshy
从文字或图片制作 3D 模型，并进行贴图、重拓扑与骨骼绑定。适合概念资产试做；绑定 API 主要适用于四肢清晰的双足人形，使用前需检查模型结构。
- Hyper3D Rodin
AI 3D generation plus model-edit workflows for localized edits and production handoff.
- Tripo
提供模型生成、分割、重拓扑、贴图、骨骼绑定与导出流程。适合把概念素材推进到可编辑资产；进入游戏或网页前，仍需检查面数、材质、尺寸和动画变形。
- Kaedim
结合 AI 生成与专业美术团队交付 3D 资产，适合需要风格匹配、修改沟通和批量交付的团队。选择前明确模型规格、修改范围和交付方式。
- Hunyuan3D
Open 3D model family focused on high-fidelity asset generation and self-hosted workflows.
- Otter
Meeting transcription and summary assistant for notes, follow-ups, and searchable records.
- Aider
Apache-2.0 terminal AI pair programmer; you pay your own LLM API keys—no Aider subscription.
- Cline
Open-source autonomous coding agent in VS Code: edits files, runs commands, and uses browser/MCP with per-step approval.
- LlamaIndex
RAG and data ingestion framework. Data connectors, query engines, and retrieval.
- LiteLLM
Unified API for many LLM providers. Routing, logging, and cost tracking.
- OpenRouter
Multi-model API gateway with pay-per-use and routing. Great for demos and multi-model tests.
- Hugging Face Inference Endpoints
将模型部署为托管推理接口，按需要配置硬件和扩缩容。适合需要独立模型服务的应用；上线前仍需验证模型效果、访问权限、延迟与运行成本。
- Replicate
Run models via API or deploy your own. Popular for AI demos and production.
- Canva
Template-first design platform for slides, social posts, and marketing visuals with AI-assisted creation.
- Stitch
Google Labs 的 UI 设计画布，可与 Stitch Agent 实时迭代；生成预览仍需独立验证响应式、无障碍和数据行为。
- Figma Slides
Presentation creation and presenting tool inside the Figma workflow.
- Gamma
AI-native presentations, sites, and docs with fast prompt-to-deck generation; optional developer API.
- Beautiful.ai
AI presentation software focused on professional slide consistency.
- Pitch
Team presentation platform with AI deck generation and collaboration.
- Figma Sites
Design, prototype, and publish websites directly from the Figma workflow.
- OpenClaw
MIT-licensed personal AI assistant you run yourself, with bridges to chat apps (WhatsApp, Telegram, Slack, Discord, Signal, iMessage, etc.) and optional coding-agent handoffs.
- Apple MLX
Apple silicon-first machine learning framework for local model inference and tuning.
- Figma Weave
在节点画布中组合生成模型与图像编辑步骤，适合复用品牌风格、比较素材变体和批量制作。先用少量样本确认效果，再扩展到多种尺寸与场景。
- Google Flow
结合文字、参考素材和首尾画面制作视频，并编辑片段与组织场景。适合分镜和视觉叙事试做；生成前确认所选模型支持的功能、积分消耗及账号地区权限。