Open source process automation: 9 workflow engines compared

Open source process automation means running a workflow engine whose source code you can read, change and host yourself. You get control over where your data lives and you pay no license fee. In return, your team runs the servers, applies the security patches and fixes the engine when it breaks. The right engine depends on who builds the workflows and how much of that work your team can take on.

Related reading: open source workflow automation platform, workflow automation platform, n8n alternatives, Activepieces vs n8n, low code workflow automation tools, CodeWords integrations.

All license facts below were checked on the projects' own pages and repositories on 1 November 2026.

Key Takeaways

• Open source engines fall into three groups: visual builders, code-first orchestrators, and data and developer platforms.
• Check the license before you choose. n8n is source-available, not open source, and Camunda 8 needs a paid license for production use.
• Activepieces (MIT) and n8n are the quickest to self-host for business workflows. Temporal, Conductor and Airflow suit engineering teams.
• The real cost of self-hosting is the team time it takes to run the engine, not the license.
• If you don't need self-hosting, a hosted platform moves that work off your team.

What is open source process automation?

Open source process automation is software that runs business or data workflows and publishes its source code under a license that lets you use, change and share it. You can host it on your own servers. The license decides what you may do with it, so "open source" on a website is only the start of the check.

The Open Source Initiative defines open source as a license with no limits on who may use the software or for what. Some popular tools publish their code but limit commercial use. Those are called source-available.

Why do teams choose open source for process automation?

Teams choose open source for three reasons: control over data, freedom to change the code, and cost at high volume.

• Data control. You decide where data lives and who can reach it. This matters in healthcare, finance and government, where data rules are strict.
• Freedom to change the code. You can add custom steps, private connectors or new run patterns. A hosted tool limits you to its plugin system.
• Cost at high volume. Hosted tools often charge per task or per run. A self-hosted engine charges nothing per run, so the cost moves to servers and team time.

Three risks balance those reasons:

• Upkeep. Security patches, upgrades, database changes and monitoring are now your team's job.
• Project health. Open source projects can slow down, split or change license. Check who funds the project and how active it is.
• Paid features. Single sign-on, roles and audit logs often sit in a paid edition.

Which open source process automation engines are worth comparing?

The table below compares 9 engines in three groups. Each one is covered in its own section after the table.

Engine	Group	License	How you build	Source
n8n	Visual builder	Sustainable Use License (source-available)	Visual flowchart, code nodes	n8n docs
Activepieces	Visual builder	MIT (community edition)	Visual flowchart	Activepieces docs
Automatisch	Visual builder	AGPL-3.0 (community edition)	Trigger and action steps	Automatisch on GitHub
Temporal	Code-first orchestrator	MIT	Code (Go, Java, Python, TypeScript and more)	Temporal on GitHub
Conductor OSS	Code-first orchestrator	Apache 2.0	JSON or code	Conductor OSS
Kestra	Code-first orchestrator	Apache 2.0	YAML	Kestra
Apache Airflow	Data and developer platform	Apache 2.0	Python	Apache Airflow
Prefect	Data and developer platform	Apache 2.0	Python	Prefect
Windmill	Data and developer platform	AGPLv3 core	Scripts in several languages	Windmill on GitHub

Visual builders

1. n8n. The most widely used self-hostable visual workflow tool. You build on a flowchart and add JavaScript or Python in code nodes. It also has AI agent nodes and an AI workflow builder. n8n uses its own Sustainable Use License, which n8n calls fair-code. It allows internal business use but limits reselling or hosting n8n for others, so it is source-available, not open source (n8n docs, 2026).

Best for: teams that want a visual builder they host themselves. Watch for: the license if you plan to embed or resell it, and database and worker tuning as volume grows.

2. Activepieces. A visual builder with a simpler interface than n8n and a permissive license. The community edition is MIT, so you may host, change and resell it (Activepieces docs, 2026).

Best for: teams that want the simplest self-hosted visual builder under a permissive license. Watch for: a smaller app library than the largest builders.

3. Automatisch. A trigger-and-action tool that feels like Zapier, for teams that want the same model on their own servers. The community edition is AGPL-3.0.

Best for: simple two-step workflows with data kept in-house. Watch for: a small community and limited support for complex branching.

Code-first orchestrators

4. Temporal. A durable execution engine. Workflows are code, and Temporal keeps them running through crashes, restarts and network failures. It handles long-running jobs, retries and timeouts. The server is MIT-licensed.

Best for: engineering teams with workflows that must finish, such as payments or order handling. Watch for: a steep learning curve and a cluster to run. There is no visual builder.

5. Conductor OSS. The continuation of Netflix Conductor after Netflix handed the project to the open source community. It coordinates workflows across many services from a central control plane, with workflows in JSON or code. Apache 2.0.

Best for: teams with many microservices that need one place to coordinate them. Watch for: it expects people with distributed systems experience.

6. Kestra. An orchestrator where you define workflows in YAML and run them on schedules or events. It covers data, infrastructure and business workflows. Apache 2.0.

Best for: teams that want workflows as readable config files under version control. Watch for: some team features sit in the paid edition.

Data and developer platforms

7. Apache Airflow. The long-standing orchestrator for data pipelines. You write workflows in Python as DAGs (a chain of tasks with a set order). Apache 2.0, run by the Apache Software Foundation.

Best for: data teams with scheduled ETL jobs, model training and batch work. Watch for: it suits scheduled batch work more than real-time events, and setup takes effort.

8. Prefect. A Python-first orchestrator. A workflow is a normal Python function with @flow and @task decorators. Local testing is easier than in Airflow. Apache 2.0, with a paid cloud dashboard.

Best for: Python developers who want orchestration with less setup than Airflow. Watch for: it focuses on data work, not general business processes.

9. Windmill. A developer platform for scripts, workflows and internal apps. It supports Python, TypeScript, Go, Bash and SQL. The core is AGPLv3, with a separate license for enterprise features.

Best for: teams that build internal tools and workflows in several languages. Watch for: AGPLv3 asks you to share changes if you offer a modified version over a network, so check it with your legal team.

A note on Camunda. Camunda was a common choice for BPMN process automation. Since version 8.6 (October 2024), Camunda 8 needs a paid license for production use (Camunda, 2024). The last feature release of Camunda 7 was scheduled for October 2025 (Camunda, 2025). That is why it is not on this list.

How do you choose an open source process automation engine?

Choose by who builds the workflows and what your team can run.

1. Who builds the workflows? People who don't write code need a visual builder: Activepieces, n8n or Automatisch. Developers can use Temporal, Conductor, Kestra, Airflow, Prefect or Windmill.
2. What does the license allow? If you plan to resell or embed the engine, rule out source-available and AGPL engines unless your legal team approves.
3. What kind of work is it? Business processes with app steps suit visual builders. Long-running, must-finish jobs suit Temporal or Conductor. Data pipelines suit Airflow, Prefect or Kestra.
4. Who runs it? Name the person who patches, upgrades and monitors the engine. If nobody owns that job, self-hosting will slip.

How does a hosted platform compare to self-hosted open source?

Self-hosted open source gives you control and no license fee. Your team carries the servers, patches and upgrades. A hosted platform carries that work for you, and you trade some control.

CodeWords is a hosted option for teams with a recurring job and no wish to run an engine. You describe the job to Cody, our builder. Cody builds the workflow, connects your tools and deploys it, on a schedule, a trigger or a request. When the job changes, you describe the change. You own the code it wrote, on every plan, which answers the main worry teams have about leaving open source. CodeWords connects to 3,000+ integrations.

If you need the engine on your own servers, CodeWords is not the fit. Use n8n, Activepieces or one of the orchestrators above. Many teams use both: self-hosted engines for data that must stay in-house, and a hosted platform for the rest.

See how it works →

FAQ

Is n8n open source?

No, n8n is source-available. It uses the Sustainable Use License, which n8n calls fair-code. You can read the code and use it for internal business work, but the license limits reselling or hosting n8n for others (n8n docs, 2026).

Which open source automation tool is easiest to self-host?

Activepieces and n8n are the easiest to self-host for business workflows. Both run from a single Docker container for a small setup. Temporal, Conductor and Airflow need more parts, such as a database, workers and a scheduler.

Can open source automation tools run AI steps?

Yes, several open source automation tools can run AI steps. n8n and Activepieces include AI and agent steps. Airflow, Prefect and Kestra can run model calls and machine learning pipelines as tasks in a workflow.

Is open source automation more secure than a hosted platform?

Open source automation is not more secure by default. You can read the code and keep data on your own servers, which helps. You also become responsible for patches, access rules and server security, which a hosted platform handles for you.

Conclusion

The best open source process automation engine is the one your team can run and keep up to date. Check the license first, then match the engine to who builds the workflows. If you would rather describe the job and let someone else run the engine, try CodeWords.