Open Source Workflow Automation Tools: 7 Options and Their Licences

Searching for open source workflow automation tools returns a list that mixes things that barely belong together: a Zapier-style app connector, a Python data-pipeline scheduler, an IoT flow editor and a durable execution engine for microservices. They are all “workflow automation,” and picking the wrong category is a much bigger mistake than picking the wrong tool within a category.

There is also a licensing trap. Several popular “open source” automation tools are source-available or have proprietary enterprise directories. That rarely matters for internal use, and matters a lot if you embed the tool in a product or run it for clients.

This guide sorts seven well-known tools into the job they are actually good at, states each licence as published in its repository (checked October 2026), and ends with a picker.

First, pick the category

Before comparing tools, decide which of these problems you have:

1. App-to-app business automation. “When a deal closes in the CRM, create an invoice and post in Slack.” Builders are ops people. Needs many SaaS connectors and a visual editor.
2. Scripts and internal tools. “Run this Python script on a schedule with a form in front of it.” Builders are engineers. Needs good code ergonomics and a UI for non-engineers to trigger jobs.
3. Data pipelines. “Every night, extract, transform and load into the warehouse.” Builders are data engineers. Needs scheduling, backfills, dependencies between tasks.
4. Durable application workflows. “An order process that waits days for payment, retries partners, and must never lose state.” Builders are backend engineers. Needs durable execution inside application code.
5. Event and device flows. “Wire MQTT sensors, HTTP calls and dashboards together.” Needs lightweight runtime and flow-based programming.

If you are unsure which you are in, our explainer what is a workflow engine walks through the differences.

The tools at a glance

Tool	Category	Licence (repository, Oct 2026)	Builder
n8n	App-to-app automation	Sustainable Use License (source-available); .ee. files under enterprise licence	Visual + code
Activepieces	App-to-app automation	MIT, except ee directories under a separate licence	Visual
Node-RED	Event and device flows	Apache 2.0	Visual flow editor
Windmill	Scripts and internal tools	AGPLv3, with proprietary enterprise code; clients Apache 2.0	Code + flows
Temporal	Durable application workflows	MIT	Code (SDKs)
Apache Airflow	Data pipelines	Apache 2.0	Python DAGs
Kestra	Data and ops orchestration	Apache 2.0	YAML + UI

n8n

Use it for: app-to-app automation by technical ops teams who want code steps and self-hosting.

n8n has a large connector library, a visual editor, JavaScript/Python code nodes and AI nodes. It is the most capable self-hostable alternative to Zapier and Make.

Licence caveat: n8n is not OSI open source. Its Sustainable Use License allows use and modification for your own internal business purposes or non-commercial use, and distribution only free of charge for non-commercial purposes. Files with .ee. in the name require an enterprise licence. Internal ops use is fine; reselling or offering it as a service to clients is where you need to talk to n8n.

Running it yourself is real work; our guide to self-hosting n8n covers the production setup.

Activepieces

Use it for: Zapier-style automation that non-developers can build, with a permissively licensed self-hosted core.

The repository’s core is MIT-licensed, with enterprise features in packages/ee and related ee folders under a separate licence. Connectors (“pieces”) are TypeScript packages, so adding a missing integration is approachable for a developer.

Trade-off: a smaller connector catalogue than n8n or Zapier. Check that your exact apps and triggers exist before committing.

Node-RED

Use it for: event-driven flows, IoT, MQTT, home and industrial integrations, and lightweight HTTP glue.

Apache 2.0-licensed and governed under the OpenJS Foundation. It runs comfortably on small hardware, and its flow-based editor is excellent for wiring events together.

Trade-off: it is not built around SaaS business apps. You can call any HTTP API, but you will be building OAuth flows and pagination yourself that n8n or Activepieces give you out of the box.

Windmill

Use it for: turning scripts (Python, TypeScript, Go, Bash, SQL) into scheduled jobs, flows and internal UIs.

Windmill’s repository states that backend and frontend are AGPLv3, with enterprise-flagged code under a proprietary licence, and client libraries under Apache 2.0. It is fast, developer-friendly, and generates input forms from script signatures, which makes it useful for ops tooling.

Licence caveat: AGPLv3 has network-use obligations if you modify it and offer it to users over a network. For internal use without modifications this is typically straightforward; get legal review if you plan to embed it.

Temporal

Use it for: durable, long-running workflows inside your application — order processing, onboarding, payment flows — where losing state is unacceptable.

MIT-licensed. You write workflows as code with Temporal’s SDKs; the server persists every step so workflows survive crashes and can sleep for days. For the model itself, see durable execution explained.

Trade-off: it is an engineering platform, not an ops tool. There are no drag-and-drop SaaS connectors, and running the server yourself is a serious undertaking.

Apache Airflow

Use it for: batch data pipelines with task dependencies, schedules and backfills.

Apache 2.0, the long-standing default for data engineering teams, with a large provider ecosystem for warehouses and cloud services. Pipelines are Python DAGs.

Trade-off: it is designed for scheduled batch work. Using it for real-time, event-triggered business automation (a webhook from your CRM) is fighting the tool.

Kestra

Use it for: declarative orchestration across data and ops tasks, defined in YAML with a UI.

Apache 2.0. Flows are YAML, which suits teams that want orchestration definitions reviewed in Git, and its plugin ecosystem covers databases, cloud services and scripts.

Trade-off: YAML-first will feel heavy for business users used to visual builders.

A licence decision tree for ops teams

Licences only matter in relation to how you plan to use the tool. Three questions cover most cases:

1. Will only your own employees use it, for your own business? Every tool on this list works. Even source-available licences like n8n’s are written to allow internal business use.
2. Will you modify it and let people outside your company use it over the network? AGPLv3 (Windmill) requires you to make your modified source available to those users. Permissive licences (MIT, Apache 2.0) do not.
3. Will you embed it in a product you sell, or run it as a paid service for clients? This is where n8n’s Sustainable Use License and any ee / enterprise directories become a blocker without a commercial agreement. Permissive cores (Activepieces, Node-RED, Temporal, Airflow, Kestra) are the safer starting point — but check that the features you rely on are not in the enterprise-licensed parts.

When in doubt, have legal read the actual LICENSE file in the version you deploy. Licences do change between versions.

What most people get wrong

• Equating “self-hosted” with “open source.” n8n is self-hostable but source-available. Windmill is open source but AGPL with proprietary enterprise code. Read the licence file, not the marketing page.
• Using a data orchestrator for business automation. Airflow and Kestra can call APIs, but you will rebuild OAuth, webhooks and connectors that app-automation tools give you.
• Underestimating operations. Every tool here needs a database, upgrades, backups and monitoring when self-hosted. Budget people-hours, not just servers; see build vs buy: the workflow question.
• Picking one tool for everything. It is normal and healthy to run Airflow for data, Temporal in the product and n8n or Activepieces for ops.

How to choose in five steps

1. Write down your top five workflows and classify each into the categories above.
2. Shortlist one or two tools for the dominant category.
3. Read each shortlisted tool’s LICENSE file and check it against how you will use it (internal, embedded, client-facing).
4. Deploy on a throwaway server with Postgres and run your most complex workflow end-to-end, including a deliberate failure.
5. Estimate monthly operating hours honestly, then compare with managed options.

When a managed platform is the better call

Open source tools give you control and no licence fees, but you take on uptime, security and upgrades. If your workflows are business-critical and nobody owns the infrastructure, a managed platform is usually cheaper in total. Automatite is one option; our n8n alternatives page compares managed and self-hosted choices side by side, and the workflow builder page shows how we approach code steps and versioning.

FAQ

Is n8n open source?

Not under an OSI-approved licence. n8n uses its Sustainable Use License, which permits internal business and non-commercial use and restricts commercial redistribution. It is often described as “fair-code” or source-available.

What is the best open source alternative to Zapier?

For non-developers, Activepieces is the closest in feel, with an MIT-licensed core. For technical teams, n8n is more capable but source-available rather than open source.

Is Airflow a workflow automation tool?

It is a workflow orchestrator for data pipelines. It excels at scheduled batch jobs with dependencies, not at event-driven app-to-app automation.

Which open source workflow tools are fully permissive?

Of the tools here, Node-RED, Airflow and Kestra are Apache 2.0, Temporal is MIT, and Activepieces’ core is MIT with separately licensed enterprise directories.