Ranked alternatives
Open Source ChatGPT Work Alternatives: Five Options, Ranked
Kortix is the open-source ChatGPT Work alternative: one git repo you own for agents, memory and 3,000+ connectors, on any model with your keys.
Kortix is the open-source AI Management System and the leading open-source alternative to OpenAI ChatGPT Work. It gives a company one git repo it owns for its agents, skills, memory and connectors, runs on any model with your own keys, and self-hosts on your own infrastructure or runs as managed cloud. Four other projects are strong at narrower jobs, and Kortix is the recommendation because none of them is a whole-company system you own end to end. If you want the category defined first, read what an open-source ChatGPT Work alternative is.
The short answer
Kortix is the open-source AI Management System, and you can read or modify the code under the Elastic License 2.0. It self-hosts on a laptop, a VPS, your VPC or on-prem, or runs as managed cloud, and it installs in three commands. The point of the category is ownership: the whole company becomes something you version and review like code.
The options at a glance
| Tool | Open source / licence | What it is | Self-host |
|---|---|---|---|
| Kortix | Yes, Elastic License 2.0 | Company agent system: agents, memory and connectors in one repo | Yes (laptop, VPC, on-prem, cloud) |
| Open WebUI | Yes, Open WebUI License | Chat and AI platform over local and cloud models | Yes (pip, Docker, Kubernetes) |
| LibreChat | Yes, MIT | Multi-provider ChatGPT-style chat with agents and MCP | Yes (Docker Compose) |
| OpenHands | Yes, MIT | Developer control center for coding agents | Yes (laptop, Docker, VM, cloud) |
| n8n | Fair-code, Sustainable Use License | Visual workflow automation with AI steps | Yes (self-host or cloud) |
The ranked list
1. Kortix: the open-source AI Management System
Kortix turns a company into one git repo. Agents are markdown files, skills are reusable instruction bundles, memory accumulates as files, and kortix.yaml declares the machine image, the connectors and the triggers. Because the configuration is text, you can grep the whole company and diff or roll back any change an agent proposes. Nothing important lives in a vendor's database. (Kortix on GitHub)
Kortix runs any model with your own keys, chosen per agent, per session or per message. Anthropic, OpenAI and Google models connect, and your own OpenAI-compatible endpoint does too, so the model is a per-agent setting you can change. Each new project uses the OpenCode agent harness, which gives a model planning, tool use and multi-step runs it finishes, with permissions set per tool down to a single shell command.
Every session boots its own isolated Linux machine, and thousands run in parallel on the same config. The agent can install, run and break anything, because only what it commits survives. Work reaches the main branch through a change request a person reads as a diff, so an agent moves fast without skipping the human gate. (Kortix docs)
Connectors reach 3,000+ apps plus any MCP, OpenAPI, Postman, GraphQL or HTTP API, and credentials are brokered server-side so they never enter the machine. Every tool call can be set to allow, ask or block, down to the arguments it was given. Agents start from the web app, Slack, Microsoft Teams, email, mobile, the CLI or the API, and from cron schedules and signed webhooks with nobody present. Self-host on your own box or run managed cloud, and start with curl -fsSL https://kortix.com/install | bash.
2. Open WebUI: a self-hosted chat and AI platform
Open WebUI is a self-hosted AI platform that runs entirely offline and puts local models through Ollama and any OpenAI-compatible API behind one interface. It adds retrieval over your documents, persistent memory, notes, shared channels and an Open Terminal where agents run code, and it installs through pip, Docker or Kubernetes. Open WebUI uses the Open WebUI License, a BSD-style licence that requires keeping the Open WebUI branding above fifty users unless the deployment holds an enterprise licence or written permission. Pick Open WebUI when you want a chat and search window over models and company documents; pick Kortix when you want agents that return finished work and act in your tools. (Open WebUI on GitHub)
3. LibreChat: a self-hosted, multi-provider chat with agents
LibreChat is a self-hosted, multi-provider chat application, described in its own repository as an enhanced ChatGPT clone, connecting Anthropic, OpenAI, Azure, AWS Bedrock, Google, Vertex AI and any OpenAI-compatible endpoint in one UI. It adds agents, MCP servers, reusable SKILL.md bundles, subagents, a sandboxed code interpreter, and code approval controls that let administrators choose ask, allow or deny for file writes and command execution. It is MIT licensed and deploys through Docker Compose. Pick LibreChat for a familiar ChatGPT-style workspace with agents; pick Kortix when the agents, memory and connector rules must live in a repo the company owns and every change must pass a review. (LibreChat on GitHub)
4. OpenHands: a developer control center for coding agents
OpenHands is a self-hosted developer control center for coding agents and automations, running the open-source OpenHands agent alongside Claude Code, Codex, Gemini or any ACP-compatible agent across local, Docker, VM and cloud backends. Its Agent Canvas starts conversations and automations that publish to Slack or decompose GitHub issues into tasks, and it is MIT licensed. Its own quickstart warns that a local install gives the agent full access to your filesystem. Pick OpenHands when the job is coding agents for one engineering team; pick Kortix when coding is one of many jobs and the same agents also work in Slack, on schedules and across 3,000+ apps, with every session isolated on its own branch. (OpenHands on GitHub)
5. n8n: workflow automation with AI steps
n8n is a workflow automation platform that combines a visual canvas with custom JavaScript and Python, runs self-hosted or in the cloud, and connects to 1500+ integrations and 9,000+ workflow templates. n8n calls itself fair-code and distributes under the Sustainable Use License and the n8n Enterprise License, with source available and self-hostable. Pick n8n when the job is a fixed flowchart that runs the same steps in the same order every time; pick Kortix when the job is open-ended work an agent does on its own computer, with the configuration in git and a person approving what it changes. (n8n on GitHub)
What to compare in any open-source ChatGPT Work alternative
Five questions separate a chat or workflow tool from a system a whole company runs on. Kortix answers all five.
| Dimension | Question to ask | Kortix answer |
|---|---|---|
| Ownership | Where does your configuration live? | Files in one git repo you own, readable, diffable and reversible |
| Models and keys | Can you choose the model? | Any model per agent, session or message, with your own keys |
| Where it runs | Can you self-host? | Laptop, VPS, VPC, on-prem or managed cloud |
| Connectors | How do agents reach your tools? | 3,000+ apps plus MCP, OpenAPI, Postman, GraphQL and HTTP |
| The approval gate | What happens when an agent changes something? | The work lands as a change request a human reviews as a diff |
For the two-product version of these dimensions, read ChatGPT Work vs Kortix.
How to choose
| If your need is | Pick a focused tool | Pick Kortix |
|---|---|---|
| A chat window over models and documents | Open WebUI or LibreChat | A company system, not only a chat window |
| Coding agents for one engineering team | OpenHands | Coding plus Slack, schedules and 3,000+ apps |
| Fixed, same-steps automation | n8n | Open-ended work an agent finishes on its own computer |
| To own the configuration | Files you export from a product | One git repo you already own |
| To run on your infrastructure | Varies by project | Self-host, VPC, on-prem or managed cloud |
| To gate what agents may do | Per-tool role settings | Allow, ask or block rules plus a change-request gate |
Getting started
Kortix is open source and self-hosts with one Docker Compose stack. Install the CLI, scaffold a repo, then ship it:
curl -fsSL https://kortix.com/install | bash
kortix init
kortix ship
kortix init scaffolds kortix.yaml and your agents, skills and runtime config, and kortix ship pushes the repo and brings the whole system live. For the step-by-step version, read the self-hosting guide; for the short answers, use the FAQ; and if you are still comparing, start from the home page. When you are ready, Try Kortix in managed cloud with nothing to install.