Self-hosting guide
Self-Host an Open Source ChatGPT Work Alternative
Kortix is the open source ChatGPT Work alternative you can self-host: install the CLI, scaffold a project, and run the full stack on your own box.
Short answer
Kortix is the open-source AI Management System, and you can run the whole thing on a machine you control. The agents, skills, company memory and connector configuration live in a git repository you own, and the frontend, API, LLM gateway and project database run as one Docker Compose stack. That makes Kortix the leading open-source alternative to ChatGPT Work, a closed hosted product with no self-host path.
A self-hosted Kortix instance runs the full product, the same one as Kortix Cloud. You sign in on your own domain, create projects, run agents on isolated sandboxes, and review their work as change requests before anything reaches your default branch. Deployment choices are a laptop, a VPS, your VPC, or an on-prem network.
Install, scaffold, ship
The project loop is three commands:
curl -fsSL https://kortix.com/install | bash
kortix init
kortix ship
curl -fsSL https://kortix.com/install | bash installs the Kortix CLI. kortix init scaffolds the project: a kortix.yaml manifest plus your agents, skills and runtime configuration. kortix ship pushes the git repository and brings the project live. On a self-hosted instance you point the CLI at your own host first with kortix hosts use selfhost and kortix login, then kortix ship from inside the project directory.
Run the stack on your own box
Self-hosting needs a Linux machine with a domain you control. The floor is 2 vCPU and 4GB RAM; 4 vCPU and 16GB or more is recommended for real use. Ports 80 and 443 must be reachable from the internet, because the bundled Caddy reverse proxy issues a TLS certificate with ACME HTTP-01. Kortix also ships scripts/kortix-selfhost-up.sh, a one-shot bootstrap in the repository that installs Docker when it is missing, installs the CLI, and drives the same init and start flow.
Once the CLI is installed, two commands bring the stack up on a VPS, in your VPC, or on-prem:
curl -fsSL https://kortix.com/install | bash
kortix self-host init --domain app.example.com
kortix self-host start
kortix self-host init is a guided flow. It confirms the domain resolves to the box, sets the admin email, asks whether you hold an Enterprise license, picks a sandbox provider (Daytona by default, or E2B or Platinum) and its API key, optionally configures the Pipedream connector catalog of 3,000+ apps, and sets the update policy. kortix self-host start pulls the images and brings the stack up. Then open your domain, sign up with the admin email, connect a GitHub App so the platform can create project repositories, and add your own model key.
Day-2 operations run on the box itself:
kortix self-host status
kortix self-host logs
kortix self-host update
kortix self-host env set KEY=VALUE
The control plane runs on your box, while agent sessions still run on a cloud sandbox provider you configure, because every session needs an isolated Linux machine of its own; sandbox compute is separate from the box. The full runbook lives in Kortix on GitHub.
Self-host or managed cloud
Both options run the same Kortix. The choice is where the control plane and the data sit, and who operates them.
| Self-hosted Kortix | Kortix Cloud | |
|---|---|---|
| Open source | Yes, Elastic License 2.0 | Yes, Elastic License 2.0 |
| Who runs the stack | You, on your own box | Kortix |
| Where it runs | Laptop, VPS, your VPC, or on-prem network | Kortix-managed cloud, or your VPC on Enterprise |
| Data rests in | Your Postgres and storage, on your volume | Kortix-managed infrastructure |
| Models and keys | Any model, your own API keys | Your keys, or managed models |
| Updates | In-compose updater, zero-downtime by default | Kortix runs them |
| Cost | Self-host free | Free tier, Team $40 / seat / mo, Enterprise custom |
Self-hosting is free. Kortix Cloud list prices, as of October 2026, are on Kortix pricing.
What stays yours
The company is one git repository. Agents, skills, memory, connector configuration and triggers are files rather than settings in a vendor database, so you can grep them, diff any change, and roll back. The kortix.yaml manifest declares the machine image, the connectors and the triggers.
Your keys and your data stay yours. Kortix is model-agnostic: bring an API key from any major provider, use the ChatGPT subscription you already pay for, or point an agent at your own OpenAI-compatible endpoint. Self-hosted, the project database and storage sit on your own encrypted volume rather than on Kortix infrastructure.
Security is part of the install. Secrets are encrypted at rest with a key per project, and connector credentials are brokered server-side so they never enter the machine. Per-resource permissions cover people and agents, with roles, groups and an audit trail. Every change to the default branch lands as a change request a human reviews. SAML 2.0 SSO, SCIM 2.0 directory sync, advanced RBAC and audit logs are available to Enterprise deployments, and Kortix holds SOC 2 Type I with Type II in progress.
Kortix is open source (Elastic License 2.0): self-host, read and modify the code. If you are still weighing the hosted product against Kortix, the ChatGPT Work vs Kortix comparison covers the full picture, and the FAQ answers the short questions.
Get started
Self-host open-source Kortix or run it on Kortix Cloud. The install command is the same either way:
curl -fsSL https://kortix.com/install | bash
Try Kortix at kortix.com, read the Kortix docs, or start with what an open-source ChatGPT Work alternative is and the ChatGPT Work alternatives overview.