n8n

Think of n8n as a node canvas where every step of an automation, from a webhook trigger to a database write, is a box wired to the next, with JavaScript or Python code nodes for anything the prebuilt blocks miss. Its AI layer maps LangChain ideas onto nodes, so an agent step can be handed a chat model from OpenAI, Anthropic, Google, or a local Ollama server, plus memory, tools, vector stores, MCP connections, and evaluation runs. More than 1,500 connectors and a large library of shared templates cover common business apps. Self-hosting means one Docker container or an npm install, with SQLite by default, PostgreSQL and a Redis queue mode for heavier loads, and an AI starter kit that adds Ollama, Qdrant, and Postgres. Read the license closely: the Sustainable Use License lets you run and modify n8n for internal business or personal use, but not resell it or host it as a service for others, and files marked .ee need a paid enterprise key. The free Community edition leaves out SSO, Git-based environments, and workflow sharing. Best for teams that want agent workflows tied into existing systems without building a backend.