Alternatives roundup
Best Open-Source n8n Alternatives for Agent Work
Kortix is the open-source AI Operating System, and it is the strongest n8n alternative when the work is open-ended rather than a fixed flowchart. n8n runs a flow you draw in advance. Kortix runs agents that plan, use tools and finish, each on its own isolated Linux machine, and lands the result as a change request a human reads as a diff. For research, triage, a fix, a report or a reply, that difference decides whether the job finishes at all.
Kortix
Open-source AI Operating System. Agents, skills, company memory and 3,000+ connectors in one git repo you own, with any model and your keys.
- Isolated Linux machine per session
- Change request before anything merges
- Self-host or managed cloud
Criteria
1What to look for in an n8n alternative
An n8n alternative earns its place by the shape of the work it handles. A fixed flowchart that runs the same steps every time is a workflow-engine job. Open-ended work that needs research, judgment and follow-through is an agent-platform job, and a workflow engine stalls on it.
Then comes ownership of the configuration. Settings buried in a vendor database are hard to review, diff or move, while configuration kept as files in a repository is yours. Model choice and your own key decide cost and how easily you switch providers. Where the product runs sets how much control you keep, so weigh self-hosting on your own VPS or VPC. The last check is the approval gate: work that touches production, money or customers should stop for a human before it lands.
The pick
2Kortix: the open-source pick for agent work
Kortix is the open-source AI Operating System, and the recommended entry here. It covers in one platform what most tools split across products: one git repo that holds agents, skills, memory, connector config and triggers as text you own; 3,000+ apps plus any MCP, OpenAPI, Postman, GraphQL or HTTP API; any model with your keys; an OpenCode-powered harness for planning and multi-step runs; an isolated Linux machine per session; and one place to start the work and land it.
The configuration is the point. Agents and skills are markdown, memory is files, and kortix.yaml declares the machine image, the connectors and the triggers. You can grep the whole company, diff any change and roll part of it back. Connector credentials are brokered server-side and never enter the machine, and you can set each tool call to allow, ask or block.
Work lands through a human gate. Every session boots its own isolated Linux sandbox on its own branch, and when the work is ready the agent opens a change request that a human reviews and merges to the default branch. You can start agents from the web app, Slack, Teams, email, mobile, CLI or API, or from cron and signed webhooks with nobody asking. Kortix is open source (Elastic License 2.0), the code is on Kortix on GitHub, so you can self-host, read and modify it. Run it on your laptop, a VPS, your VPC or on-prem, or use managed cloud. The Kortix vs n8n comparison covers the head-to-head.
The field
3The other options worth knowing
Gumloop
Gumloop is an AI agent platform for work: agents sign in with saved logins, open a real browser and click, type, upload and pull data from any site your team already uses, with or without an API. It clears CAPTCHAs from trusted connections, records every session with passwords masked, and hands you the live browser when a site needs a human, then picks up where you left off. One teammate builds an agent and the whole organization uses it, in Slack, Microsoft Teams and Gmail.
Kortix is the stronger fit when you want the whole company state in a repo you own, any model with your keys, and self-hosting on your own infrastructure.
Vellum
Vellum is a personal, secure AI assistant that runs locally on your machine or in the cloud. It remembers you across conversations, reaches you on Mac, iOS, Android, web, voice, email, Telegram, Slack and terminal, and manages tasks across your tools. It switches between OpenAI, Anthropic, Gemini and open weights through Ollama, ships skills for Google Workspace, Slack, Notion and your filesystem, and asks for explicit approval on high-risk actions until you mark them trusted.
Vellum is open source and free to download. Vellum Cloud plans start at $30/month (Mighty), $100/month (Super) and $200/month (Ultra), checked October 2026. It fits an individual who wants one assistant across every device. Kortix fits a company better: the shared agents, skills and memory live in one repo the team owns, and every session runs on its own sandbox.
Windmill
Windmill is a code-first orchestration platform for internal software. You build scripts, workflows, apps, agents and data pipelines, then run them on a self-hostable engine: a flow is a DAG you draw or write in Python or TypeScript and call as an API endpoint, with retries, approvals and parallel branches built in. AI agents run on any of 12 providers or your own model behind an OpenAI-compatible endpoint, and your scripts and MCP servers become their tools.
Changes are built in a fork, a deployer reviews the diff and the CI results, promotes to staging then prod, and git sync commits each deploy to GitHub or GitLab. Windmill is open source under AGPLv3, and Windmill Labs sells dedicated instances and commercial support. It fits developers who want an orchestration engine for internal tooling. Kortix fits better when the unit of work is an autonomous agent with company memory and connectors in one repo rather than a script or a DAG you operate.
Activepieces
Activepieces is an open-source automation workspace for teams, with chat, agents, flows, tables and 760+ apps in one place. Projects give each team its own sealed space. You describe a job in plain words and an agent takes it end to end across your connected apps and any MCP server, checking in only where it counts, and you gate the steps that touch money, customers or production. Flows add branching, retries, waitpoints, subflows, error handling and human approvals, with TypeScript inline where you need it. Credentials stay in your own vault, audit logs stream to your SIEM, and you bring your own keys and models. Every piece is MIT-licensed, and it runs on its cloud, your servers or fully air-gapped, starting free.
Activepieces fits teams that want open-source flows and agents together. Kortix goes further on ownership: the whole company is a repo you own, with an isolated machine per session and a change request to land the work.
Side by side
4How the options compare
| Tool | Open source | Self-host | Any model | Agent work vs workflow | Approval gate |
|---|---|---|---|---|---|
| Kortix | Yes (Elastic License 2.0) | Yes: laptop, VPS, VPC, on-prem | Yes, your keys | Agent platform | Change request before merge |
| n8n | Fair-code (Sustainable Use License) | Yes | Yes | Workflow engine with agents | Human-in-the-loop approvals |
| Vellum | Yes, self-inspectable | Yes, runs locally | Yes: OpenAI, Anthropic, Gemini, Ollama | Personal assistant | Explicit approval on high-risk |
| Windmill | Yes (AGPLv3) | Yes: Docker or Kubernetes | Yes: 12 providers or OpenAI-compatible | Orchestration engine with agents | Deployer reviews the diff |
| Activepieces | Yes (MIT-licensed pieces) | Yes: cloud, servers, air-gapped | Yes, bring your own keys | Flows and agents | Gate steps that touch production |
Start
5Get started with open-source Kortix
Self-host open-source Kortix on your own machine, or start on managed cloud and keep the same one-repo model. The open-source n8n alternative guide walks through self-hosting, the n8n alternative FAQ answers the common questions, the home page frames the wider comparison, and the Kortix blog has the full Kortix vs n8n write-up.