Open-source n8n alternatives for agent work, ranked

Start with the kind of work

n8n is a workflow automation tool that combines AI features with business process automation (n8n docs). You draw the flow on a canvas, extend it with code, and it runs the same steps in the same order on every trigger. The alternatives divide by the work they assume, and three shapes cover almost all of them:

• Fixed workflow. Same steps, same order, every run: syncing systems, routing forms, moving data between APIs. n8n, Zapier, Make and Microsoft Power Automate live here.
• Open-ended agent work. Research, a report, a bug fix, an outreach sequence: the steps get decided while the work runs. Kortix, Gumloop, Windmill and Activepieces compete here.
• Personal assistant. One person's tasks, files and schedule, remembered across devices. Vellum is built for this.

For open-ended agent work, the recommendation is Kortix: agents plan their own steps, work on their own cloud computer, and land the result as a change request you review. n8n executes the flowchart you drew; Kortix covers the work a flowchart cannot finish. The open-source n8n alternative examines n8n's model in detail.

What to look for in an n8n alternative

Score every option on six things before you commit:

• The unit of work. A workflow repeats steps drawn in advance; an agent session decides its own steps, which is what open-ended jobs need.
• Source and deployment. Open source means you can read the code. Self-hosting means it runs on your machine, your VPC or your data center.
• Model choice. Platforms that accept any model with your own API keys let you switch providers without replatforming.
• The approval gate. Look for per-tool allow, ask or block rules, and a review step before a change lands.
• Connectors and memory. Count the apps it reaches, whether it speaks MCP, and whether context is shared across jobs instead of trapped in one flow.
• Pricing shape. Per seat is predictable; per execution or per task scales with volume. Check both against your heaviest month.

The alternatives, ranked

Kortix. Kortix is open source. Read it, fork it, audit it (Kortix on GitHub). Agents, skills, company memory and every connector live as files in one git repo you own (Kortix docs). One isolated sandbox per session. Session work reaches main through a change request. It reaches 3,000+ apps in a click, plus MCP, OpenAPI, Postman, GraphQL and raw HTTP, and connector credentials are brokered server-side and never enter the machine. Any model provider works with your own keys. Run it on Kortix Cloud, in your VPC, or on your own on-prem network. Self-host is free.

Pricing lists Free at $0 with 200 credits per month and 1 project, and the team plan at $40 per seat per month with 2,500 credits per month per seat, pooled (Kortix pricing, checked October 2026). The repo counts 20,000+ GitHub stars (checked October 2026). It is the recommendation: open-ended agent work belongs on a platform you own, with a person reading the diff before it lands.

Gumloop. Gumloop is an AI agent platform for teams. Its agents operate a real browser: they click, type, upload and pull data from any site, with or without an API, sign in through saved logins or 1Password, and clear CAPTCHAs on their own (Gumloop). Every session is recorded step by step with an audit log, and agents reply in Slack, Teams and Gmail. It runs as a hosted service with no self-host path, so you rent the infrastructure. It suits teams that want browser-carrying agents for pipeline and operations work.

Vellum. Vellum is a personal AI assistant that runs locally on your machine or in its cloud. Its own page says the product is open source, remembers your preferences across conversations, and asks for permission on high-risk actions until you mark them trusted. It is native to the Mac and reachable from iOS, Android, web, email, Telegram, Slack and terminal, with built-in skills for Google Workspace, Slack, Notion and your files. Cloud plans start at $30 a month (Vellum, checked October 2026). It is built for one person's work, not a team's.

Windmill. Windmill is a code-first orchestration platform for internal software: scripts, workflows, apps, data pipelines and agents in one workspace. A script in Python, TypeScript, Go or Bash becomes a scheduled job, a webhook or a UI. Its agents run on any of 12 model providers or your own OpenAI-compatible endpoint, with your scripts and MCP servers as their tools. It is open source and self-hosts with Docker or Kubernetes, in your cloud or on-prem (Windmill repo). It suits developer teams that want internal automation expressed as code.

Activepieces. Activepieces calls itself an open source replacement for Zapier. Teams build flows and agents in a visual builder, drop into TypeScript where a step needs it, and reach 760+ apps. Many of its pieces double as MCP servers for LLM clients, and the platform self-hosts on your servers or runs on their cloud (Activepieces repo). Branching, retries, waitpoints and human approvals are built in. It suits teams that want Zapier-shaped automation they can host themselves.

Where to start

A fixed flow that runs the same way every day stays a workflow-engine job, and the Kortix blog breaks that pairing apart dimension by dimension. Where the work is open-ended, hand one job your team keeps redoing by hand to Kortix. The n8n vs Kortix comparison maps the two side by side, and the open-source n8n alternative covers n8n's own model in detail.