中文教程
中国大陆用户如何使用 Codex：安装、可用性与模型接入指南
了解中国大陆地区的官方服务可用性、Codex CLI 安装方法，以及第三方模型接入前需要核实的条件。
The considered AI directory
Good tools.
Better decisions.
Discover AI tools by the work you want to do. Compare the possibilities, with the limits in view.
中文资源
安装与入门教程
中文教程
Claude Code 安装教程：macOS、Windows 和 Linux 入门
根据 Anthropic 官方文档安装 Claude Code，完成版本检查、账号登录和首次项目操作。
中文教程
Meta Muse 桌面版安装教程：Mac 下载与首次设置
通过 Meta 官方页面下载 Muse for Mac，了解首次登录、权限设置和使用前检查。
Find your next tool
Explore the atlas.
More filters
Audience tags describe product positioning, not nationality or guaranteed regional access.
Source-check dates, with names breaking ties. Documentation dates; test scope is stated on each tool page.
Compare up to 3 tools
Automation
Activepieces
A visual automation platform with a self-hosted community edition and optional cloud hosting.
For Operators building reviewable automation
Key limitation
Self-hosting and connected services require independent operating and access checks.
Docs checked 2026-11-06
Automation / Building & coding
LangGraph
A real model-free LangGraph/SQLite checkpoint test: approval, restart recovery and the boundary between replay and duplicate writes.
For Developers who need explicit state and review steps
Key limitation
A framework requires engineering and does not automatically guarantee safe external actions.
Docs checked 2026-11-06
Building & coding
Aider
A terminal coding assistant for working with selected files in a Git repository while keeping changes visible for review.
For Developers making small, testable repository changes
Key limitation
Source context may leave the device with remote models.
Docs checked 2026-10-28
Research & office
AnythingLLM
A document and assistant workspace offered as desktop software, a self-hosted server and managed cloud, with edition-specific capabilities.
For Teams working with an approved document collection
Key limitation
Permissions and features vary between desktop, server and cloud.
Docs checked 2026-10-28
Building & coding
Bolt
A browser-based AI builder for JavaScript web applications, with project editing, hosting and GitHub integration.
For Teams prototyping a JavaScript web application
Key limitation
Check usage costs and an independent code-and-data backup.
Docs checked 2026-10-28
Building & coding
Claude Code
A repository-focused coding agent for editing files, running commands and reviewing changes across terminal and editor workflows.
For Developers automating a bounded repository task
Key limitation
Command permissions and billing depend on the selected mode.
Docs checked 2026-10-28
Sales research / Research & office / Automation
Clay / Claygent
A hosted go-to-market workspace for structured account research and enrichment, with separate platform Actions and Data Credits to assess.
For Sales and go-to-market operations teams that need structured account research and controlled enrichment.
Key limitation
A generated account attribute can be unsupported or stale and should be reviewed before CRM changes or outreach.
Company research for sales · The documented research and enrichment steps fit a specific prospect question; validate evidence, data coverage and credit spend.
Docs checked 2026-10-28
1.67M estimated site visits · Sep 2026 · Semrush
Building & coding
Cline
A coding agent across editor and terminal workflows, with permissions, model access and operating spend that need explicit choices.
For Engineers supervising repository tasks
Key limitation
Permissions depend on mode, tools and settings.
Docs checked 2026-10-28
Research & office / Automation / Building & coding / Video
Coze / 扣子 (coze.cn)
Coordinate people, agents and project files in the China-facing product, while checking connected-tool permissions and the exact team agreement.
For Chinese-speaking teams evaluating a project workspace for coordinated research, creation and delivery.
Key limitation
Shared project context and multiple agents do not establish independent review or bounded permissions for connected accounts.
Docs checked 2026-10-28
Building & coding
Cursor
An AI code editor with repository search, agent-assisted edits and command execution.
For Maintainers working through an existing codebase
Key limitation
Privacy settings do not turn connected inference into local processing.
Docs checked 2026-10-28
Research & office / Automation
Genspark
A hosted AI workspace for turning research and office tasks into documents, slides and other deliverables, with credits and feature-specific quotas to check.
For Knowledge workers and small teams producing research-backed office deliverables.
Key limitation
Generated documents and slides still need source checks, calculation checks and export-format review.
Docs checked 2026-10-28
17.08M estimated site visits · Sep 2026 · Semrush
Customer support / Sales research
Gorgias
An ecommerce support platform with an AI Agent, configurable actions and human handover; evaluate it against the store policies and channels you actually use.
For Ecommerce support teams that can define store policies, permitted actions and reliable human escalation.
Key limitation
Correct support depends on current store knowledge and carefully scoped action permissions, with a tested human handover path.
Ecommerce support triage · Commerce integrations and human handoff make it a candidate for order-related support; verify supported systems and escalation rules.
Docs checked 2026-10-28
Research & office / Building & coding
Haystack
A Python framework for building explicit retrieval and agent pipelines, intended for developers who will own evaluation and deployment.
For Developers building explicit retrieval pipelines
Key limitation
Authentication, evaluation and operation remain your work.
Docs checked 2026-10-28
Video / Design & commerce
HeyGen Video Agent
A hosted prompt-to-video workflow with scene planning and editable output; check the exact plan, language feature and rights before production use.
For Creators and teams producing reviewed explainer, marketing or onboarding video with controlled assets.
Key limitation
A generated render still requires factual, pronunciation, timing and likeness-rights review before use.
A narrated product walkthrough · Its documented video workflow is a candidate for a script-led walkthrough; verify likeness rights, language coverage and revision limits.
Docs checked 2026-10-28
Research & office / Building & coding
Kimi Agent
Turn a research brief into inspectable files, with separate checks for source accuracy, shared usage allowances and the agreement covering your work.
For Researchers and office users who can review source evidence and editable deliverables.
Key limitation
A generated file or working preview does not establish factual accuracy, backend permissions or the right service agreement for customer work.
Docs checked 2026-10-28
20.57M estimated site visits · Sep 2026 · Semrush
Research & office
LibreChat
A self-hosted conversation workspace for multiple model endpoints, with a broader operating stack than a standalone chat page.
For Teams using several approved model providers
Key limitation
Plan database services and per-user access controls.
Docs checked 2026-10-28
Research & office / Building & coding
LM Studio
A proprietary desktop application for local model experiments, with current offerings that also include optional cloud access.
For Builders exploring local models
Key limitation
App and model terms differ; hardware still limits capacity.
Docs checked 2026-10-28
Building & coding
Lovable
An AI web app builder for interfaces, databases and authentication, with editable project code and GitHub export.
For Builders validating a small web workflow
Key limitation
Code export alone does not prove data or deployment portability.
A small web application · A hosted builder with project-code export suits a bounded workflow; prove the database permissions and independent recovery path before handoff.
Docs checked 2026-10-28
81.46M estimated site visits · Sep 2026 · Semrush
Design & commerce / Video
Lovart
Develop product and campaign visuals from an approved brief, while protecting product accuracy and checking model-specific commercial rights.
For Designers, marketers and merchants who can approve product accuracy, typography and commercial rights.
Key limitation
Visually convincing assets can still alter product facts, packaging or text and require human approval.
A product-asset brief · Its design-agent and canvas workflow fits a coordinated visual brief; product accuracy, image rights and revisions still need human checks.
Docs checked 2026-10-28
6.17M estimated site visits · Sep 2026 · Semrush
Research & office
Manus
Use parallel research for a repeatable item set, then verify coverage, citations, account visibility and cost before publishing the result.
For Analysts and independent operators comparing many items against an inspectable evidence schema.
Key limitation
Research coverage, entity matching and source support still require review before results guide publication or decisions.
Source-led company research · Its documented parallel research workflow is a useful candidate for comparable company questions; check every cited result and task-credit cost.
Docs checked 2026-10-28
40.42M estimated site visits · Sep 2026 · Semrush
Research & office
Open WebUI
A self-hosted chat interface for teams choosing their own model endpoints, with document features and an important branding-license condition.
For Teams operating an internal chat interface
Key limitation
Branding terms apply; remote inference remains remote.
Docs checked 2026-10-28
Building & coding / Automation
Replit Agent
A hosted development agent that plans, builds and refines applications in a shared workspace; evaluate generated behavior and usage billing separately.
For Founders, product teams and developers building a small application with inspectable behavior.
Key limitation
Generated code and agent-run tests do not establish production security, correct permissions or an adequate recovery plan.
Docs checked 2026-10-28
22.72M estimated site visits · Sep 2026 · Semrush
Research & office
Skywork
Assess a research-to-document workspace through citations, editable exports and the agreement that permits your intended use.
For Researchers and office teams assessing citation-backed documents and editable presentations.
Key limitation
The linked access license and current account entitlements need confirmation before relying on Skywork for commercial deliverables.
Docs checked 2026-10-28
1.7M estimated site visits · Sep 2026 · Semrush
Research & office
Vane (formerly Perplexica)
A self-hosted search-and-answer application combining web retrieval with configurable models and citations to inspect.
For Researchers inspecting retrieved web evidence
Key limitation
Search coverage and individual citations need verification.
Docs checked 2026-10-28
Design & commerce / Video
ComfyUI
Arrange generative media operations as a node graph with explicit model and workflow choices.
For Builders creating repeatable media workflows
Key limitation
Model and custom-node rights and hardware requirements vary.
Docs checked 2026-11-06
Building & coding
Continue
Use an open source coding assistant inside an editor, with configurable models and project context.
For Developers choosing their own model workflow
Key limitation
Review provider handling and every proposed code change.
Docs checked 2026-11-06
Automation / Research & office
Dify
Build AI applications with workflow orchestration, knowledge retrieval and an API-facing publishing layer.
For Teams building an internal AI workflow
Key limitation
Edition permissions, license conditions and model costs vary.
An internal AI workflow · The workflow and knowledge features make it a candidate for a bounded shared process; evaluate edition permissions and license conditions first.
Docs checked 2026-11-06
Automation
Flowise
Connect components in a visual workspace for chat, retrieval and agent workflows.
For Teams prototyping visual AI workflows
Key limitation
Review access controls and connected-tool permissions.
Docs checked 2026-11-06
Automation / Building & coding
Langflow
Prototype Python-oriented AI flows and expose a selected workflow through an application interface.
For Python teams composing AI pipelines
Key limitation
Custom components execute code and need secret isolation.
Docs checked 2026-11-06
Research & office / Building & coding
Ollama
Run models through a local application and connect a repeatable inference step to your own workflow.
For Local model prototyping
Key limitation
Model rights, hardware and inference location need separate checks.
Docs checked 2026-11-06
No matching tools.
Try a different task or clear the filters.
Your shortlist
Compare up to 3 tools. Your selection stays when filters change.
| Compare |
|---|