Open source automation tools cover two related but distinct jobs: connecting apps and services into no-code workflows, and building the logic pipelines behind AI agents and chatbots. Activepieces and Huginn represent the classic integration-platform-as-a-service (iPaaS) model, similar to Zapier or IFTTT: triggers from one service (a new row in a spreadsheet, an incoming webhook, an RSS feed) fire actions in another, chained together in a visual builder without writing code. Huginn is older and more developer-oriented, running agents that scrape, filter, and forward data on a schedule. Activepieces is closer to a modern Zapier alternative, with a large library of pre-built app connectors and a drag-and-drop flow editor.
Langflow, Dify, and Flowise sit in a newer part of the category: visual builders for large language model workflows. Each lets a developer or non-developer wire together prompts, retrieval steps, tool calls, and conditional logic into an agent or chatbot without writing a full application from scratch. Flowise and Langflow are node-based canvases built on LangChain-style abstractions. Dify adds a more complete application layer on top, including a hosted API, prompt versioning, and a dataset feature for retrieval-augmented generation. Novu covers a narrower but common need: multi-channel notification workflows (email, SMS, push, in-app) triggered from application code, with a workflow editor for approvals, digesting, and channel fallback.
What to look for depends on the job. For connecting SaaS tools without code, check the connector library size and whether the trigger model is polling-based or webhook-based, since webhooks are faster and cheaper to run. For LLM workflow builders, check which model providers and vector databases are supported out of the box, and whether the finished flow can be exposed as an API endpoint for use inside a real application. For notification workflows, check delivery provider support (SendGrid, Twilio, Firebase) and whether user preference management is built in. All six tools can be self-hosted with Docker, which matters for teams running workflows over private data or internal systems that should not pass through a third-party SaaS platform.