n8nn8n
n8n 是「自动化优先」的低代码工作流平台——AI 是它众多节点里的一类,不是它的身份。 它最容易被误读的地方是许可:源码公开,但它是 fair-code(Sustainable Use License),不是 OSI 开源。
n8n is an automation-first low-code workflow platform — AI is one node type among many, not its identity. The most misread part is the license: source is public, but it is fair-code (Sustainable Use License), not OSI open source.
适合与不适合When to use it, when not to
适合:跨系统自动化是主轴、要把 AI 嵌进既有流程、要自托管掌控数据与凭据。 不适合:把 n8n 包进产品卖给客户(商业协议线);纯代码编排(去 OpenAI Agents SDK 等 framework 大类); 开箱团队共享(Community 版工作流仅创建者可见)。
A fit for: cross-system automation as the main axis, embedding AI nodes into existing flows, self-hosting for data control. Not a fit for packaging n8n into your product and selling it to clients (requires a commercial agreement); pure code orchestration (go to the framework category); out-of-the-box multi-tenant SaaS (Community has no sharing).
中文生态China availability 国产 / 国内可直接用 / 备案状态 · 快照domestic / direct access / CAC filing · snapshot 2026-10-21
- 出品属地Origin
- 国外出品foreign-made n8n GmbH(柏林)。n8n GmbH (Berlin).
- 国内使用Direct access
- 大陆可直连direct OK docs.n8n.io / support.n8n.io 直连实测 200(2026-10-21)。docs.n8n.io / support.n8n.io measured 200 direct (2026-10-21).
- 备案状态CAC filing
- 未见于备案名单not in filing list:未见于网信办公示名单(2026-10-21 检索);该产品未面向大陆公众提供服务,备案不适用。not in the CAC public registry (searched 2026-10-21); the product does not serve the mainland public, so filing does not apply.
备案信息以网信办《生成式人工智能服务已备案信息》公示名单为准(按批次滚动更新,本站只抄编号不解读);「大陆可直连」是本站从大陆网络直连实测的结果,不是厂商承诺。GitHub 类站点当日实测有波动,口径写在依据里。Filing status is quoted from the CAC public registry (updated batch by batch; this site copies the number and does not interpret it). "Direct access" is this site's own reachability measurement from a mainland network, not a vendor promise. GitHub-hosted sites fluctuated on the measurement day — the note says so.
固定坐标系Fixed coordinate system
8 个维度dimensions, 与同分区其他对象逐项可比each comparable item by item against others in the same section.
- 模型与开放条件Model & openness
平台本身不带模型,AI 节点经各厂商凭据接外部 LLM(OpenAI、Anthropic、Google 等经 LangChain 节点族与专用节点);具体节点覆盖清单未逐条核验,以官方 nodes 文档为准。 自托管时模型凭据在你自己的实例上。
The platform ships no model itself; AI nodes reach external LLMs via per-vendor credentials (OpenAI, Anthropic, Google etc. through the LangChain node family and dedicated nodes); the node coverage list has not been verified item by item — the official nodes docs are authoritative. When self-hosted, model credentials live on your own instance.
- 运行位置Where it runs
两条官方形态:自托管(Docker/npm,Community 版免费无执行上限)与 n8n Cloud 托管订阅。 自托管队列模式(worker 扩展)包含在 Community 内,multi-main 高可用属付费档 (官方 choose-n8n 口径,2026-10 第三方转述+官方 docs 对照)。
Two official forms: self-hosted (Docker/npm, free Community edition with no execution cap) and n8n Cloud managed subscription. Self-hosted queue mode (worker scaling) is included in Community; multi-main HA belongs to paid tiers (official choose-n8n framing, 2026-10 third-party paraphrase cross-checked against official docs).
- 本地文件Local files
自托管实例跑在你自己的机器/服务器上,工作流数据、凭据、执行日志都落在自有环境; 二进制数据外部存储(S3 等)属付费档。Cloud 形态数据在 n8n 云端。
A self-hosted instance runs on your own machine/server — workflow data, credentials and execution logs all land in your own environment; external storage for binary data (S3 etc.) is a paid tier. In the Cloud form, data sits in n8n's cloud.
- 关机后的任务Tasks after shutdown
工作流由实例执行:自托管实例常开即可承载定时与 webhook 触发;官方给 Cron/队列机制。 「电脑关了还能跑」取决于你把实例跑在哪——平台不替你保管关机后的执行。
Workflows are executed by the instance: a continuously running self-hosted instance carries scheduled and webhook triggers; the official docs provide Cron/queue mechanisms. "Keeps running when your laptop is off" depends on where you run the instance — the platform does not hold executions for you after shutdown.
- 工具与扩展Tools & extensions
核心是集成节点生态(官方称数百种集成,未逐条核验);MCP 在官方许可 FAQ 里被点名 (「通过 n8n 编辑器、你的界面、n8n API、MCP 或 AI agent 配置工作流」需商业授权场景之一) ——MCP 是官方支持的使用面。
The core is the integration-node ecosystem (officially "hundreds of integrations", not verified item by item); MCP is named explicitly in the official license FAQ ("configuring workflows through the n8n editor, your UI, the n8n API, MCP or an AI agent" is one of the commercial-license scenarios) — MCP is an officially supported usage surface.
- 上下文与记忆Context & memory
工作流内的数据变换与子流程传递是主轴;跨工作流的持久记忆机制未逐条核验 (未知就说未知)。凭据与执行数据按实例持久保存。
In-workflow data transformation and sub-workflow passing are the main axis; cross-workflow persistent memory mechanisms have not been verified item by item (unknown means unknown). Credentials and execution data persist per instance.
- 权限与限制Permissions & limits
自托管权限边界=你的实例边界;Community 版工作流/凭据仅创建者与实例所有者可访问 (共享属付费档)。企业 SSO/RBAC 属 Business/Enterprise licence key 解锁。 许可边界比权限边界更关键:SUL 三条限制见 license 字段。
Self-hosted permission boundary = your instance boundary; in Community, workflows/credentials are accessible only to their creator and the instance owner (sharing is a paid tier). Enterprise SSO/RBAC unlocks via Business/Enterprise license keys. The license boundary matters more than the permission boundary: SUL's three restrictions see the license field.
- 适合什么任务Best for
适合:跨系统自动化为主轴、要把 AI 节点嵌进既有流程、要自托管掌控数据。 不适合:要把 n8n 包进自己产品卖给客户(需商业协议);要纯代码编排(去看 framework); 要开箱多租户 SaaS(Community 无共享)。
Fits: cross-system automation as the main axis, embedding AI nodes into existing flows, self-hosting to control data. Does not fit: wrapping n8n into your product to sell to customers (needs a commercial agreement); pure-code orchestration (go to framework); out-of-the-box multi-tenant SaaS (Community has no sharing).
头号误解Top misconception
- 把「源码可得」当「开源」——SUL 是 fair-code,内部业务用途之外有硬限制Reading “source available” as “open source” — SUL is fair-code with hard limits beyond internal business use
- 把 n8n 托管给客户收费——SUL 明确不允许,需 n8n Embed 商业协议Hosting n8n for clients for money — explicitly not allowed under SUL, needs the n8n Embed agreement
- 以为 Community 能团队协作共享工作流——共享/项目/SSO 属付费档Assuming Community supports team workflow sharing — sharing/projects/SSO are paid tiers
- 把 AI 能力当独立产品——n8n 主轴是自动化,AI 是节点之一Treating AI capability as a standalone product — n8n's axis is automation; AI is one node type
未知项清单Known unknowns
- AI 节点覆盖的 LLM provider 全清单(官方 nodes 文档未逐条核验)
- 跨工作流的持久记忆机制(未核验到官方文档)
- 官方「数百种集成」的确切数量(未逐条核验)
- The full LLM-provider catalog covered by AI nodes (official nodes docs not verified item by item)
- Cross-workflow persistent memory mechanisms (no official documentation found)
- The exact count behind the official "hundreds of integrations" (not verified item by item)
证据来源Evidence sources
判断可回到以下一手源复核。本站核验日Every judgement can be rechecked against the primary sources below. Verified on 2026-10-30,内容更新日content updated on 2026-10-21。
| 类型Type | 名称Name | 链接Link |
|---|---|---|
| docs | n8n · Community license 官方页(SUL 三条限制、.ee. 文件例外、与旧 Apache+Commons Clause 沿革)n8n · Community license official page (SUL's three restrictions, the .ee. file exception, and the lineage from the old Apache+Commons Clause) | https://docs.n8n.io/n8n-community-license |
| docs | n8n 帮助中心 · Which license do I need(内部业务用途判例、MCP 被点名、客户不能开编辑器)n8n help center · Which license do I need (internal-business-use case rulings, MCP named explicitly, customers may not open the editor) | https://support.n8n.io/article/can-i-use-your-license-for-my-use-case |
| repo | n8n-io/n8n · 仓库(208,219★,最近推送 2026-10-21,GitHub API 核验)n8n-io/n8n · repository (208,219★, last push 2026-10-21, verified via GitHub API) | https://github.com/n8n-io/n8n |
| changelog | n8n · Release notes(版本迭代记录入口)n8n · Release notes (version iteration records entry point) | https://docs.n8n.io/release-notes/ |
实测记录Measured runs
本档未跑工作流搭建实测;结论只覆盖平台能力与权限边界。
This site has not completed measured runs. The closest test protocol is the code-side protocols/framework/_protocol.md (the tool-calling and credential-boundary sections are reusable; no workflow-platform-specific protocol exists on this site yet).
许可是它最该先读的一页The license is the page to read first
官方 Community license 页给 SUL 三条限制(要点转述):只能用于自有内部业务或非商业个人用途; 分发必须免费且非商业;不得移除许可与版权声明。文件名含 .ee. 的部分归 Enterprise License。
官方帮助中心进一步给了判例式口径:你在自己实例上替客户跑工作流、客户只拿结果——允许; 客户能进编辑器或经你的界面/API/MCP/AI agent 配置工作流——需要商业协议(n8n Embed)。
The official Community license page states SUL's three restrictions (key points paraphrased): use only for your own internal business or non-commercial personal purposes; redistribution must be free of charge and non-commercial; the license and copyright notices may not be removed. Files whose names contain .ee. fall under the Enterprise License.
The official help center adds case-law-style guidance: running workflows for a client on your own instance while the client only receives the results — allowed; the client entering the editor or configuring workflows through your UI/API/MCP/AI agent — requires a commercial agreement (n8n Embed).
相关条目Related entries
- OpenAI Agents SDKOpenAI Agents SDK代码优先 vs 画布优先对照:同一个「让 AI 调工具干活」,一边写代码一边拖节点。Code-first vs canvas-first contrast: the same "let AI call tools to do work" — one side writes code, the other drags nodes.
本页由This page is generated from the ai-handbook 数据层生成(CC BY 4.0)。data layer (CC BY 4.0).
方法论与坐标系定义见仓库内For the methodology and coordinate definitions, see METHODOLOGY.md.