Self-host an open-source ChatGPT Work alternative
You can run Kortix, the open-source ChatGPT Work alternative, on a Linux box you control. Three commands install the CLI, initialize the instance, and start the stack. Your data, your files, and your model keys stay on hardware you own.
Three commands
$ curl -fsSL https://kortix.com/install | bash$ kortix self-host init --domain kortix.example.com$ kortix self-host startFrom the Kortix self-hosting docs. Point DNS at the box and open ports 80 and 443 first. Then check kortix self-host status, logs and doctor while the stack starts.
What a self-hosted Kortix runs
Kortix is the open-source AI Operating System, and a self-hosted instance runs the whole control plane inside your own network. One Docker Compose stack carries the frontend, the API, the LLM gateway, and the Supabase distribution. Your accounts, projects, repos, secrets, connectors, policies, and audit log sit on disk you own.
Agent sessions run on a separate sandbox provider. The default provider is Daytona; Platinum and E2B are also supported. The instance must be reachable so the sandbox can call back to it.
What the first run asks
kortix self-host init asks six things and no model key. The CLI generates every port, URL, password, signing key, and Compose default itself, and writes the whole instance to one directory. The six things are:
- A domain you point at the box, or a Cloudflare tunnel for evaluation.
- An admin email, which grants platform admin.
- Whether to unlock SAML 2.0 single sign-on, SCIM 2.0 directory sync, custom roles, groups, and audit read.
- The sandbox provider (Daytona, Platinum, or E2B) and its key, the one credential the stack cannot start without.
- Pipedream credentials for the app catalog, optional and skipped by default.
- Whether to update nightly automatically.
GitHub and your model key are not asked at init. GitHub connects in the dashboard at Settings → Git, and models are BYOK in the app, where you add a provider key or sign in with a subscription. See models and your keys for the model setup.
What self-hosting still needs
A self-hosted Kortix instance needs the internet. kortix self-host start pulls the stack images from docker.io, so the box needs outbound access to pull them. The instance must also be reachable from the sandbox: a session runs on a remote provider and calls back to your stack over its domain or tunnel.
Self-hosting versus ChatGPT Work
ChatGPT Work is OpenAI's hosted agent for longer, multi-step work and finished deliverables, running in OpenAI's cloud on eligible paid ChatGPT plans. Kortix runs work of the same kind on hardware you control, with the configuration stored as files in a git repo you own. Self-host is free. The managed cloud is a separate option, with seats and pooled credits; see pricing for the plan shape, and ChatGPT Work vs Kortix for the full side by side.
Self-hosting questions.
Can I run a self-hosted Kortix on a VPS?
Yes. The same stack runs on a laptop, a VPS, or a cloud server, and a domain is one environment variable to set.
How does a self-hosted instance get updated?
Every instance updates itself automatically. Pin an exact version with kortix self-host update --tag <version>, or turn the updater off with --auto-update off.
How do I back up a self-hosted Kortix instance?
Copy two data directories and the instance .env file: volumes/db/data, volumes/storage, and .env, which holds every secret and signing key. There is no separate backup service.
Does a self-hosted instance include managed models?
No. A self-hosted instance has no managed model lineup. You connect the providers you already pay for, and every model call routes through the gateway running inside your own instance.
Run Kortix on hardware you control.
Self-host is free, and the whole control plane comes up from three commands.
Open source · Any model, your keys · Self-host, VPC, or on-prem