Is OpenHands open source, and what does open source actually get you?

Is OpenHands open source?

OpenHands is open source. Its own site calls it an open source AI coding agent platform, its documentation calls the browser client behind it the open-source control center for agent conversations, and the repositories are public: OpenHands on GitHub is where you confirm it by reading the code. That answer is not in dispute, and the access it gives is real for a team that wants to inspect what an agent does before it does it.

The inspection earns its keep, because a coding agent runs commands on a real machine. The repository's own install notes warn that the direct local install runs the agent server on the machine you install it on, with full access to that filesystem, and offer a Docker route that gives each conversation its own runtime. Open source puts that warning in front of you instead of in a support thread: you can read the code, build it, and decide how much of a machine to hand over.

What open source covers, and what it leaves open

Reading the source tells you what a program does. It says less about the setup that surrounds it: whose keys the agent uses, where the files it touches live, who owns the configuration, and who approves a change before it lands. Those questions decide how much of a company an agent can carry, and they are answered by the repository and the review path, not by the source on its own.

Open source code, open source operating system

An open-source coding agent is a program you can read and run. 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. In a Kortix project those agents, skills, memory, tools, triggers and connectors are files in the repo, beside kortix.yaml at the root, and Kortix either hosts the repo or you connect your own. Every session runs one agent in its own sandbox, on a branch named after the session id. The agent commits, pushes and opens a change request, and a person with the project.gitops.merge permission merges it. Merge is default-deny for agents, so an agent merges its own change request only when its grant in kortix.yaml includes that permission.

That is the distinction that decides how far a team can take it. OpenHands' code answers what the agent does with a codebase. Kortix answers that, and who owns the instructions, the memory and the approval. The first is a repository you can read. The second is a repository that is also the company, versioned and diffable, with changes reviewed the way code is reviewed. Kortix is the leading open-source alternative to Claude Cowork and ChatGPT Work for teams that want the second.

Self-hosting, keys and data

Run it on Kortix Cloud, in your VPC, or on your own on-prem network. Self-host is free, and the kortix self-host commands install and run the full stack on your own infrastructure. Models stay your choice: any model provider with your own keys, or the ChatGPT plan you already pay for. Connector credentials are brokered server-side and never enter the machine. A secret can be egress-enforced, where the sandbox holds a handle and Kortix substitutes the real value outside the sandbox only on approved HTTPS hosts; environment is the default exposure and egress enforcement is experimental. Every action is recorded. Reading, exporting and streaming the audit log depend on the plan. The self-hosting page has the commands and what the first run asks for.

OpenHands reaches a similar place from the other direction. Agent Canvas starts with one npm install or one docker run, runs on your laptop by default, and can point at local, Docker, VM or cloud backends. The company sells hosted execution, collaboration and budget controls through OpenHands Cloud, and self-hosted Kubernetes deployments inside your own VPC through OpenHands Enterprise.

What to check before you choose

Six questions separate a readable agent from a system a team can run on. Can you read the code? Where does the configuration live, and does it diff? Where does a session run, and what can it reach? Who approves a change before it reaches the default branch? What happens to your keys and to the data a session touches? What does each tier cost?

Kortix answers them with a repo you own, one isolated sandbox per session, merge by a person by default, and a free tier of $0 with 200 credits per month and 1 project; the team plan is $40 per seat per month with 2,500 credits per month per seat, pooled. OpenHands answers the first question plainly and leaves the rest to how you deploy it, which suits an engineering team that wants a coding-agent control center and is content to run the backend. Where the deciding factor is who owns the configuration and who approves the change, OpenHands compared with Kortix sets the two side by side, and what each one costs sits next to it.

What a project is in Kortix

Kortix and OpenHands side by side

Seven questions that separate readable agent code from a system a team can run on.

Open source questions

The four things readers ask about OpenHands, Kortix and self-hosting.