OpenHands vs Kortix: a coding control center or an open source AI Operating System

Kortix's own row for OpenHands

OpenHands vs Kortix, side by side

Nine answers decide it: what each one is, what a unit of work is, where the work runs, who approves a change and what it costs.

How the two differ in kind

OpenHands Agent Canvas is a control center for coding agents. Its repository introduces it as the self-hosted developer control center for coding agents and automations, and the unit you start there is a conversation, pointed at an agent server on your laptop, in Docker, on a VM or in a cloud backend (OpenHands on GitHub).

Kortix runs a wider set of jobs, and the Kortix docs make the scope explicit. Kortix is the open source AI Operating System: a project is a git repo with kortix.yaml at its root, and your agents, their skills, company memory, tools, triggers and connectors are files inside it. The unit of work is a session, which gets its own cloud computer and its own branch and ends by opening a change request.

The Kortix docs then read like the parts of an operating system. Sessions are the processes, the project repo is the file system, agents and skills are the programs, connectors are the drivers, triggers are the scheduler, roles, grants and connector policies are the permissions, and change requests are the updates.

That difference decides most of the comparison. Agent Canvas answers where a coding agent runs and how to automate the work around it. Kortix answers which agents do the company's work, what they can reach, what they remember, and who approves what they change.

Where OpenHands is the stronger instrument

For work that is code, Agent Canvas is the more focused tool. It runs the OpenHands agent out of the box and can drive Claude Code, Codex, Gemini CLI or any agent that speaks the Agent-Client Protocol, and you can move between local, remote and cloud backends without moving your conversations. The OpenHands Software Agent SDK is a Python and REST foundation for building agents that work with code, and the vendor documents a large-codebase path for mapping dependencies and running agents in parallel.

It also stands up in one command. npm install -g @openhands/agent-canvas starts the local stack, a Docker image does the same on a server, and each conversation can get its own Docker runtime when you want more separation. The repository is straight with you about the direct local install: it runs the agent server on the machine you install it on, with access to that machine's filesystem.

Self-hosting costs nothing, and the cloud Individual plan is listed free with a cap of 10 conversations a day (OpenHands pricing, checked October 2026). Agent Canvas is in beta, which its repository states (checked October 2026), so expect the control center to move quickly.

Where Kortix is the platform

The jobs that decide this comparison are rarely code only. A support question in Slack, a weekly report, a CRM update, a release note: Kortix runs all of them on one layer, because a session is an agent on a real cloud computer with the company's tools in reach. Connectors cover 3,000+ apps in a click, plus MCP, OpenAPI, Postman, GraphQL and raw HTTP, and a trigger starts work on a cron schedule or a signed webhook with nobody present.

Review is part of that same layer. Session work reaches main through a change request, merge is default-deny for agents, and a person with the project.gitops.merge permission merges it. Around that sit approval gates you set, per-tool rules that run down to the arguments of each call, per-resource permissions for people and agents, roles, groups and an audit trail. Slack and Microsoft Teams are live, and email is experimental.

You choose where it runs and which model it uses. Run it on Kortix Cloud, in your VPC, or on your own on-prem network, with self-hosting free. 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, and every action is recorded in the audit log.

What changes in a team's day

• The instructions become a diff. In Agent Canvas, skills sit in .agents/skills/ and repository context in AGENTS.md, with memory under ~/.openhands/memory on the machine. In a Kortix project the agents, skills, memory, triggers and connectors are files in the repo beside kortix.yaml, so changing how the company works is a change somebody reviews.
• The work stops being only code. A Kortix session can start from a Slack mention, a Teams message, a cron entry, a webhook, the CLI or an MCP client, and the agent that fixes a bug can also draft the customer reply.
• Who merges the change. Both systems land work as git review. OpenHands agents open pull requests and your team reviews them in the git host; a Kortix session opens a change request against the project's default branch, and who may merge is a grant you control.
• Where the machines come from. OpenHands needs a machine, a VM or a backend you point it at. Kortix gives each session its own cloud computer, and the kortix self-host commands install the full stack on your own infrastructure when you want it there.

Running both

They are not exclusive, and the seam is a git repository. Agent Canvas counts OpenCode among the agents it can run, and OpenCode is the harness a Kortix project runs. Skills for the coding conversations can sit in .agents/skills/, and the company's agents, skills and memory sit beside them in the same repo, because both are plain files. A workable split: engineers keep Agent Canvas pointed at the repos they live in, and the company routes cross-team work through Kortix sessions, with one git history underneath both.

Which one to pick

If the agents are meant to do the company's work, Kortix is the platform to run it on: the open source AI Operating System, one repo for agents, skills and memory, a cloud computer per session, reach into Slack, Teams and 3,000+ apps, and a change request a person merges. Kortix is the leading open-source alternative to Claude Cowork and ChatGPT Work for exactly that job. Keep OpenHands where the work is code and the control center is the point, and if you are still weighing the field, the OpenHands alternatives list covers the rest while openhands-alternative.com holds the other comparisons.

OpenHands and Kortix, the questions that decide it

Four questions decide the choice: open source, self-hosting, cost and fit.