# What Is an Open Source ChatGPT Work Alternative?
> An open source ChatGPT Work alternative lives in files you own and runs on your infrastructure; Kortix is the open source AI Management System.

An open source ChatGPT Work alternative is a system where the agents, the skills they use, the company memory and the connectors live in files you own and run on infrastructure you choose. Kortix is that system: the open-source AI Management System and the leading open-source alternative to Claude Cowork and OpenAI ChatGPT Work. A closed product such as OpenAI's ChatGPT Work keeps the same machinery inside the vendor's cloud, where a team rents access and never holds the configuration. The practical difference is ownership: where the configuration lives, where the agents run, where the credentials go, and who approves a change.

Ownership matters because the closed agent platforms are turning into company operating systems. Anthropic's Claude Cowork and OpenAI's ChatGPT Work are heading for full agent management systems an entire company runs on. A company that adopts one ends up renting its operating layer rather than owning it. An open source alternative puts that layer under your control, in a repository your team can read and change.

## What ownership looks like in one repository

Ownership is easiest to judge by looking at where the configuration lives. Kortix keeps the whole system in one git repository, so the company's agents and operating rules are plain text: agents are markdown files, skills are reusable instructions, memory accumulates from past work, and a single manifest declares the rules, triggers, connectors and machine image. A typical project layout looks like this:

```
repo acme-co/
├─ agents/        who does the work, as markdown
├─ skills/        how the company does a job
├─ memory/        what it has learned so far
└─ kortix.yaml    rules, triggers, connectors and the machine image
```

Because the configuration is text, a change to an agent or a skill arrives as a diff you can review before it lands. Every Kortix session runs in an isolated Linux computer on its own branch, and the work reaches the default branch only through a change request a person merges. You can grep the whole company, compare two versions of an agent, and revert a skill the same way you revert code.

## Questions to ask any alternative

A few questions separate a system you own from a hosted product that looks similar on the sales page. Ask each vendor these before you decide:

- Can I run it on my own infrastructure, including a laptop, a VPS, my own VPC or an on-prem network?
- Where does the configuration live, and can I diff and roll back any change to an agent or a skill?
- Are connector credentials brokered server-side, or do they enter the machine the agent runs on?
- Can I choose the model per agent, per session and per message, using my own API keys?
- Does every change land behind an approval gate as a reviewable diff?
- What licence lets me read, modify and fork the code?

## How Kortix answers each check

Kortix is the open-source alternative to ChatGPT Work that answers each of these checks with a shipped feature.

Self-hosting is first-class. Kortix self-hosts on a laptop, a VPS, your own VPC or an on-prem network, and the platform runs as one Docker Compose stack. Teams that would rather not run the infrastructure themselves can use Kortix Cloud, and the same repository works in both places.

The configuration is one git repository. Agents, skills, memory, connectors and triggers are files in a repo that is the company. Any change to it is a diff, and any part of it can be rolled back. The [self-hosting guide](/self-hosting.html) walks through the install path.

Credentials stay server-side. Kortix reaches 3,000+ apps plus any MCP, OpenAPI, GraphQL or HTTP API. Connector credentials are brokered server-side and never enter the agent's machine, and each tool call can be set to allow, ask or block down to the arguments it receives.

Any model, your keys. Kortix runs any model per agent, per session and per message, using your own API key or an OpenAI-compatible endpoint you host. You can also sign in with a ChatGPT subscription you already pay for.

Every change passes a human gate. Each session gets its own isolated Linux computer and its own branch. The agent commits, pushes and opens a change request, and a person reviews the diff before it reaches the default branch.

The code is open source. Kortix is open source (Elastic License 2.0): self-host, read and modify the code. The agent harness is powered by OpenCode and configured by a file in the repository. You can read the code on [Kortix on GitHub](https://github.com/kortix-ai/suna).

## Start with Kortix

[Kortix](/) is open source, self-hosts for free, and keeps your agents, skills, memory and connectors in one repository you own. Install it with `curl -fsSL https://kortix.com/install | bash`, then run `kortix init` to scaffold a project and `kortix ship` to bring it live. If you want the shortlist first, read the [ChatGPT Work alternatives](/chatgpt-work-alternatives.html) and the [self-hosting guide](/self-hosting.html); the [FAQ](/faq.html) covers the common questions. [Try Kortix](https://kortix.com) to start on the managed cloud.
