What is Claude Cowork, and what is the open source alternative?
Claude Cowork is Anthropic's agent for knowledge work, and the open-source alternative is Kortix, the AI Operating System you run and own. You give Cowork a goal, point it at folders and tools, and it returns finished documents for review, running on Anthropic's models and its cloud.
One repo is the company
# kortix.yamlagents: support-triage: file: agents/support-triage.md # what it does, in plain markdown connectors: all # what it may reachtriggers: - slug: daily-digest type: cron cron: '0 0 9 * * 1-5' prompt: Summarize yesterday's support tickets and open a change request with the digest.The agents and triggers are files in a git repo you own.
What Claude Cowork is
Claude Cowork is Anthropic's agent for knowledge work. It uses the same agentic architecture as Claude Code with no terminal, and instead of answering one prompt at a time it takes a goal and works it through to finished output: a spreadsheet, a deck, a memo. Anthropic describes it as the place you hand Claude real work. It runs on desktop, and on the web and mobile apps (Claude Cowork product page).
Cowork is available on paid Claude plans only. The same page lists Pro with Cowork included from $17 per month and the Team plan at $20 per seat per month, and every plan carries usage limits (checked October 2026).
How Claude Cowork works
You set the goal and point Cowork at the material. It reads the folders you connect and reaches apps through connectors. Then it plans the task, breaks it into subtasks, runs code and shell commands, and coordinates workstreams in parallel. Anthropic shows each step as it happens, so you can watch the files it opens and the tools it uses, and redirect it.
On a desktop session Cowork can also read and write your local files directly, with no upload step. When the task is done, the output lands in the session for you to review and download. Files reach Claude through the Claude Desktop app on the machine that has it open, and only for the folders you have connected.
What it runs on, and where
Cowork runs on Anthropic's models only. There is no picker that reaches another vendor's model.
For Team and Enterprise plans, sessions run in Anthropic's cloud on Anthropic's servers, in an isolated environment created for that one session, and your sessions and files are saved to your Claude account (Anthropic's get-started article). Anthropic also lets an organization run it where its data lives, through its own Amazon Bedrock, Google Cloud or Microsoft Foundry account (product page). What you cannot do is run Cowork on your own servers. There is no self-host.
Where the configuration lives
Everything you build in Cowork lives in the Claude product and your Claude account. Skills, plugins, connectors, projects and memory are configured there. Projects are workspaces with their own files, links, instructions and memory, and for a cloud session those are saved with the member's Claude account (Anthropic's Team and Enterprise article).
That configuration is not a repo you own. You cannot clone it, grep it, diff it, or roll it back with git.
What to check if you want to own the system
If the goal is a system your company owns, the row that matters is ownership: who holds the code, which model it runs, where the configuration lives, and where the work runs.
Kortix is the open-source alternative for exactly that. Kortix is the open-source AI Operating System, where your agents, their skills, your company memory and every connector live in one git repo you own, and each session runs on its own cloud computer. Run it on Kortix Cloud, in your VPC, or on your own on-prem network, with self-host free, and any model provider with your own keys.
Put the two side by side in the standings table and in Claude Cowork vs Kortix. The Kortix team also wrote about the pair on the Kortix blog.
Claude Cowork questions, answered
Does Claude Cowork run on models other than Claude?
No. Cowork runs on Anthropic's models only, and there is no picker that reaches another vendor. Kortix is model-agnostic, and runs any provider with your own keys.
Can you self-host Claude Cowork?
No. Cowork runs in Anthropic's cloud, or in your own Amazon Bedrock, Google Cloud or Microsoft Foundry account, and Anthropic documents no self-host. Kortix runs on Kortix Cloud, in your VPC, or on your own on-prem network.
Where does Claude Cowork keep the instructions you give it?
In the Claude product and your Claude account. Skills, plugins, connectors, projects and memory are configured there, and a cloud project is saved with your account. That is not a repo you own, so git cannot version it.
Is there an open source Claude Cowork alternative?
Yes. Kortix is the open-source AI Operating System built as the alternative: your agents, skills, memory and connectors live in one git repo you own, and each session runs on its own cloud computer.
What does Claude Cowork cost to start?
Cowork is included on paid Claude plans. Anthropic lists Pro with Cowork included from $17 per month, with usage limits on every plan (checked October 2026). Kortix is free to self-host with your own model keys.
Own the system your agents run on
Your agents, skills, memory and connectors in one git repo you own.
Open source · Any model, your keys · Self-host, VPC, or on-prem