Kortix vs OpenHands: an open-source company system or a coding-agent harness?

OpenHands is a strong, MIT-licensed harness for coding agents; Kortix is the open-source system for a whole company's agent workforce. Both self-host and both let you bring your own model, so the deciding question is the job. If you need agents that read and change code in your repositories, OpenHands is the specialist. If you need to own and govern the agents, skills, memory and connectors that run across a company, Kortix is the pick.

What OpenHands is great at

OpenHands is an open-source coding-agent harness and developer control center that runs agents on your laptop, in Docker, on a VM, or inside your own cloud. Its repository calls it the self-hosted developer control center for coding agents and automations, and the code ships under the MIT license (OpenHands on GitHub, checked October 2026).

• It drives the coding agents you already use. Agent Canvas runs the built-in OpenHands agent plus Claude Code, Codex, Gemini, or any agent that speaks the Agent-Client Protocol, and you switch between local, remote and cloud backends from the same frontend.
• Self-hosting is a first-class path. Run it directly on a laptop, in a Docker container per conversation, on a dedicated machine, or on a server in your company's infrastructure. The repository documents each route.
• Model choice stays yours. The project pitches bring-your-own-model and works with any LLM, including open-weight models you host yourself.
• Execution is sandboxed. The runtime builds a Docker image per project and runs shell, file and Python actions inside it before returning observations to the agent backend (runtime architecture).
• Automations reach developer tools. Agents run on a schedule or a webhook and connect to Slack, GitHub, Linear, Notion and similar services.

The project is mature and actively released: v1.24.0 shipped recently (releases; checked October 2026).

Where a coding-agent harness stops

OpenHands is built around code. Its agents read repositories, run terminals and tests, and open pull requests. That focus is the point of OpenHands, and it sets the boundary.

• The job is software. Agent Canvas is a development control center. It does not run outbound or close the books across departments.
• Memory is per conversation, not per company. Conversations and workspace files persist, but no single repository holds the company's agents, skills, memory and connector rules as versioned files a team reads and diffs.
• Execution starts on a host. The documented quickstart runs the agent server directly on your machine and warns that it then has full access to your filesystem; the Docker and VM backends isolate it. That suits a developer, and it is not one isolated machine per session across an organization.
• The tool surface is developer-shaped. Integrations center on Slack, GitHub, Linear, Notion and webhooks rather than the 3,000+ business apps a company-wide workforce touches.
• Review lands on code. OpenHands opens pull requests on engineering changes; it does not put agent work from every department behind one reviewed change request.

OpenHands does that job well, and these are the limits of the job itself.

What Kortix does differently

Kortix is the open-source AI Management System: one git repository holds the agents, skills, company memory, connector configuration and triggers, and every session runs on its own isolated Linux machine (Kortix docs). The company is the repository.

A short kortix.yaml declares how a session boots:

# kortix.yaml (excerpt; the manifest lives at the repo root)
default_agent: ops-morning
sandbox:
  image: Dockerfile
agents:
  ops-morning: { connectors: [slack, stripe] }
triggers:
  - { type: cron, agent: ops-morning, cron: '0 7 * * 1-5' }

• One repo you own. Agents and skills are Markdown, memory is files, and connector rules and triggers are config. Grep the whole company, diff any change, roll it back.
• One gate to land work. A session runs on a branch named after it, the agent commits, and the work reaches main only through a change request a human reads as a diff. Merge is deny-by-default for agents.
• Any model, your keys. Pick a model per agent, per session or per message: Claude, OpenAI, Gemini, the ChatGPT subscription you already pay for, or your own OpenAI-compatible endpoint.
• Every business tool. Kortix connects 3,000+ apps plus any MCP, OpenAPI, GraphQL or HTTP API. Credentials are brokered server-side so raw keys never enter the machine, and you set allow, ask or block per tool call.
• Every channel you work in. Start sessions from the web app, Slack, Teams, email, mobile, CLI or API, or from cron and webhooks with nobody asking.
• Deploy anywhere. Self-host on a laptop, a VPS, your VPC or on-prem, or use managed Kortix Cloud. Self-hosting is free; the Team plan is $40 per seat per month with pooled credits, and the free tier starts at $0 with 200 credits a month (Kortix pricing, checked October 2026).

Kortix is open source (Elastic License 2.0): self-host, read and modify the code. The source is at Kortix on GitHub.

Where OpenHands wins

If your problem is coding agents on your own repositories, OpenHands is the right tool for the job. It is MIT-licensed, runs any model, isolates execution in Docker, and carries a large community and a fast release cadence. A team that wants a self-hosted development control center for code, and nothing beyond it, should start there. Kortix asks you to treat the whole company as a codebase, which is more than a coding team needs.

The other name you'll hear: OpenClaw

OpenClaw is an open-source personal AI assistant that runs on your own computer and meets you in the chat apps you already use, from Discord and iMessage to Slack, Teams, WhatsApp and 20+ more (OpenClaw on GitHub). State, memory and credentials stay on your hardware, models and harnesses are swappable plugins, and there is no paid tier or hosted service. The OpenClaw Foundation stewards it as a 501(c)(3), and the code is MIT-licensed. For one person who wants a private, always-on agent, it is excellent, and we recommend it. OpenClaw is built for one person. A governed company platform is a different product, with a shared control plane for many people, least privilege and audit.

Side by side

When to pick which

For one person who wants a private assistant in their own chat apps, OpenClaw is the better start. For a coding team that wants a self-hosted development control center, OpenHands is the better start. For a company that needs one governed system whose agents, memory and connectors live in a repo it owns, Kortix is the one to run.

The wider field is in the open-source Microsoft Copilot alternative map and personal AI agents vs a company OS. To choose a platform rather than a harness, read how to pick an open-source AI agent platform.

Own the whole agent workforce, not just the coding loop. Kortix is free to start and free to self-host. Get started with open-source Kortix.