An open-source Zapier alternative for developers

Why developers outgrow the drag-and-drop builder

A hosted builder keeps your automation in someone else's database. You can click a node, but you cannot clone it, diff it or review it before it runs. Kortix is the open-source AI Operating System: your agents, their skills, your company memory, and every connector in one git repo you own, with the agents working on real cloud computers. If you have been looking for a Zapier alternative on GitHub, the repo is the point.

The unit of work is a session, not a step. You hand an agent a job in plain language, and it plans its own steps and runs them. A flow runs the steps you drew; a session takes a job you can describe but cannot draw. That difference matters when the work is research, a fix or a report whose steps you cannot list in advance.

What lives in the repo

Three kinds of files carry the work:

• agents are who does the work, one markdown persona per job.
• skills are how the company does a job, markdown plus scripts.
• memory is what the company has learned so far.

One repo holds them all, plus the kortix.yaml that declares the machine image, the connectors and the triggers. You can grep the whole configuration, diff any change, and roll any part of it back. The repo behaves the same on your laptop and in the cloud, so a session in the cloud runs the same files you tested locally.

Connectors with per-call control

The repo also holds the connectors. Kortix reaches 3,000+ apps in a click, plus MCP, OpenAPI, Postman, GraphQL and raw HTTP. Connector credentials are brokered server-side and never enter the machine, so a session calls a tool without holding the key.

You decide what each call may do. Each tool call can be set to Allow, Ask or Block, with rules down to the arguments of each call. An Ask holds the call, and the agent's turn pauses and resumes when a person approves. Approval gates you set. Off until you set them.

Any model, chosen per agent

Kortix is model-agnostic. Pick the model per agent, per session or per message, with any model provider and your own keys. Or the ChatGPT plan you already pay for. When a better model lands, you switch it in a file instead of waiting on a hosted builder's roadmap.

Each session runs on its own isolated machine

One isolated sandbox per session. Each session has its own isolated machine and branch, and session id, sandbox id and branch name are the same string. The agent can install, run and break anything, because only what it commits survives. Thousands of agents in parallel on one config, each on its own cloud computer, is the part a laptop cannot do and a fixed canvas cannot either.

Work lands through a review gate

Session work reaches main through a change request. Merge is default-deny for agents, so the company improves one reviewed change at a time. You read the diff first and decide whether it lands. Triggers run the same way with nobody present: a cron schedule or a signed webhook starts a session, and the agent opens a change request with the result.

Kortix or a flow builder

If every job you have is a fixed flow between apps, a flow builder covers it and it is the simpler choice. Choose Kortix when the job is open-ended, when you want the configuration in a repo you own, or when you want an agent that takes a real job on a real machine and hands back a change you can read. The Kortix vs Activepieces walkthrough covers that split in full.

Start with three commands: curl -fsSL https://kortix.com/install | bash, then kortix init, then kortix ship. Self-host is free, or run it on Kortix Cloud.