n8n made workflow automation approachable for engineers: a visual canvas where API calls, conditionals, and AI steps connect into integrations you would otherwise hand-code and babysit. The friction is not the tool but the license. n8n's source is available rather than open by the standard definition, governed by terms that restrict how you can offer it to others and reserve features for paid tiers, so what looks open carries commercial strings the moment you build a product on top of it. The open source alternatives below cover node-based orchestration of services, data, and AI steps under licenses without those reservations. You can run them, modify them, and embed them in your own offering without checking whether your use case crossed a line.
Dify is an open-source platform for building LLM applications from prototype to production. A single visual interface combines AI workflows, RAG pipelines, agent capabilities, and model management, so teams can design, test, and ship AI apps without stitching together separate tools.
Flowise is an open source platform for building AI agents and LLM workflows visually. Instead of writing every step in code, you assemble agents, retrieval, and orchestration logic by dragging and connecting nodes on a canvas, then run them as live applications.
Node-RED is a low-code programming tool for event-driven applications. It uses a browser-based flow editor to wire together devices, APIs, and online services, making it useful for building automation and data-handling flows without writing everything from scratch.
Windmill is an open-source developer platform for internal code: APIs, background jobs, workflows, and UIs. It turns scripts into shareable internal apps, webhooks, scheduled jobs, and flows, giving teams a code-first way to run automation and internal software from one platform.
Trigger.dev is an open source platform for building AI workflows in TypeScript. It is built for long-running tasks that need retries, queues, observability, and elastic scaling without the limits of short-lived serverless jobs.
StackStorm is an integration and automation platform for services and tools. It takes actions in response to events so DevOps and SRE teams can automate troubleshooting, auto-remediation, incident response, and deployment steps across existing systems.
Inngest provides durable functions for reliable background jobs, workflows, and AI agents without extra infrastructure. It replaces queues, state management, and scheduling so developers can build step functions and background logic that can run on serverless infrastructure, servers, or the edge.
Match the pick to your workflow style - general automation or AI orchestration.
Closest to n8n's visual automation: Node-RED Node-RED is the nearest match to how n8n works: a browser-based, low-code flow editor for wiring devices, APIs, and online services into event-driven automations, with a large library of community nodes and shareable flows. An OpenJS Foundation project under Apache 2.0, it is genuinely open, without n8n's source-available strings, and runs anywhere from a Raspberry Pi at the edge to a cloud server. Install it with npm and start it from the command line.
Best permissively licensed AI workflow builder: Flowise Flowise is a visual canvas for building AI agents, multi-agent systems, and RAG over your own documents, with drop-in nodes for third-party services and an API to embed any flow into another app. Its Apache 2.0 license is genuinely permissive, so you can modify and ship it in a product without the source-available strings that pushed you off n8n. Install it with npm or the Docker image and self-host, or start on its cloud.
Best all-in-one LLM app platform: Dify Dify goes wider than a single canvas, combining AI workflows, RAG pipelines, agent tools, a prompt IDE, and model management across many inference providers in one interface. Every app it builds ships with a matching API, so embedding it in your own backend is straightforward, and it self-hosts with Docker Compose or Kubernetes. Pick it when you want to design, test, and operate LLM apps end to end rather than only wire up automations.
Most teams leave n8n over its license, not any failing in the product, so the first thing to get straight is what you are actually replacing. The options here split by style. Node-RED is the closest to n8n's spirit - a browser-based, low-code flow editor for wiring devices, APIs, and services into event-driven automations. Windmill takes a code-first path, turning scripts in Python, TypeScript, Go, and more into jobs, workflows, and internal apps. If your workflows are mostly AI and LLM orchestration - prompts, retrieval over your own data, agents calling tools - Flowise and Dify fit that lane instead. Be honest about which half you use, because a general connector library and an AI-agent canvas are different products.
The tools differ in reach. Node-RED is a mature low-code flow editor with a large library of community nodes, and it runs anywhere from a Raspberry Pi to the cloud under the permissive Apache 2.0 license. Windmill is code-first, parsing your Python or TypeScript scripts into autogenerated UIs, flows, and webhooks, backed by Postgres and released under AGPLv3. On the AI side, Flowise is a visual canvas for agents and RAG over your own documents, also Apache 2.0, while Dify is broader still, bundling AI workflows, RAG pipelines, agent tools, a prompt IDE, and model management with an API for every app. Node-RED, Windmill, and Flowise carry standard open source licenses without n8n's source-available strings, and all four self-host so the automations run on your own infrastructure.
Whatever you choose, migrate on behavior rather than files. Treat n8n's workflow JSON export as documentation, not a portable package: rebuild each flow in the target system, re-enter credentials through its own secret store, and replace n8n-specific nodes with HTTP calls or native integrations. Run both systems in parallel, replay real payloads, and check that retries, waits, and webhook responses match before you cut over schedulers and endpoints.
Those are two different questions. n8n has used a source-available license rather than a standard OSI-approved open source one, so confirm the current terms before treating it as open source. Flowise, by contrast, is Apache 2.0, which is genuinely permissive. Switching can cut your license exposure, but you still pay for hosting, monitoring, upgrades, backups, and the time spent rebuilding workflows, so budget for effort, not just fees.
Both kinds are here. Node-RED, Windmill, StackStorm, and Inngest are general automation tools - visual flow editing, code-first jobs, and durable orchestration across many services - so they cover the SaaS glue and scheduled-job side of n8n. Flowise and Dify are AI-first, built for LLM orchestration, agents, and RAG over your own data. Match the tool to your workflows: general integration glue points to Node-RED or Windmill, while heavy AI steps point to Flowise or Dify.
Do not assume they will. n8n stores credentials separately from workflow logic, and its JSON export references credentials rather than carrying usable secrets. Build a fresh secret inventory, rotate anything stale, and re-enter values through the target system's own credential store. It is also a good moment to replace shared personal tokens with scoped service accounts.
No. Some common services map cleanly, but node behavior is rarely identical, and n8n nodes often hide pagination, rate-limit handling, field mapping, and auth details. When there is no equivalent, fall back to raw HTTP requests, a small script, or a custom integration. Validate the edge cases rather than only the happy path, because that is where a migrated workflow tends to fail quietly.
Treat them like an API migration. The new tool will usually generate different webhook paths, signing behavior, or response timing, so update the calling systems and, where possible, keep the old endpoints behind a proxy during the transition. Test retries from the sender, and pay special attention to any workflow where n8n returned data synchronously to the caller, since that timing is easy to break.
Some tools keep it, some do not. Node-RED is a browser-based flow editor, and Flowise and Dify both provide a visual canvas, so non-developers can edit automations - prioritize clear validation and safe credential handling there. Windmill and the durable-orchestration tools are more code- or config-first, which is fine when developers own the workflows and changes go through code review and automated tests instead of a drag-and-drop editor.
Choose an exit path before you need one. Favor tools that store workflows in readable files or documented database tables, expose an API, and can run jobs without a hosted control plane - the options here self-host with Docker and expose an API. Keep your workflow documentation and test payloads outside the tool, so a second migration later is an inconvenience rather than a crisis.
