ARTICLE / SEPTEMBER 26, 2026

Kortix vs OpenWork: Which Open-Source Claude Cowork Alternative Fits Your Team?

Kortix vs OpenWork compares two open-source Claude Cowork alternatives across license, execution model, security, pricing, and self-hosting.

Kortix is an open-source AI management system — a platform where a company's agents, skills, memory, and connectors live in one git repository and every agent session runs in an isolated sandbox on its own branch, landing work through a change request a human approves.

Kortix vs OpenWork is a comparison between two products that both describe themselves as open-source alternatives to Claude Cowork, but which are built around fundamentally different shapes. Kortix is a cloud-native AI management system where each agent session runs on its own isolated machine and work reaches the main branch only through a reviewed change request (source). OpenWork is a local-first desktop application, built on OpenCode, where AI agents work on your own files, with an optional source-available control plane for teams (source). This page compares the two across license, execution model, isolation, model choice, team permissions, and pricing, and names the cases where each is the wrong fit. For the wider category, see our roundup of open-source Claude Cowork alternatives.

By Kortix Blog · Published September 26, 2026 · Last updated September 26, 2026

What OpenWork is

OpenWork is a free, open-source desktop application for macOS, Windows, and Linux where AI agents do real work on your own files (source). It is built on OpenCode and positions itself as "the open-source alternative to Claude Cowork and Codex" (source). In desktop mode your files stay on your machine and prompts are sent directly to the LLM provider you choose; cloud is optional rather than required (source). OpenWork works with any model — 50+ providers, your own API keys, local models via Ollama, or ChatGPT sign-in — and lets teams share skills and MCP servers (source).

OpenWork's licensing is directory-split, in the style of GitLab. Everything outside the ee/ directory — the desktop app and core platform — is MIT licensed, which is an OSI-approved open-source license. Everything under ee/ (OpenWork Den, the organization control plane) is under the OpenWork EE License, a source-available license whose code is public so you can audit what you deploy, but which requires a subscription for production use, except that it is free for organizations up to 5 users, free to evaluate for 30 days at any size, and always free for development and testing (source).

What Kortix is

Kortix is the open-source AI management system that describes itself as "the leading open-source alternative to Claude Cowork and ChatGPT Work" (source). Rather than a desktop app, Kortix is a cloud-native system where a company's agents, skills, memory, and connectors all live in one git repository — versioned, diffable, and shared by the whole company (source). Each session runs an agent in an isolated sandbox on its own branch; when the work is ready, the agent opens a change request that a human reviews and merges to the default branch (source).

Kortix is source-available under the Elastic License 2.0 (ELv2), not an OSI-approved open-source license (source). The source is public, forkable, and self-hostable for free, but ELv2 restricts offering the software to third parties as a hosted or managed service. Kortix runs on your own infrastructure — a laptop, VPS, VPC, or on-prem network — or on Kortix's managed cloud at $40 per seat per month plus usage (source).

Kortix vs OpenWork at a glance

The table summarizes the two products on the dimensions that matter most for a build-or-buy decision. Every cell links to the primary source it is drawn from.

DimensionOpenWorkKortix
LicenseMIT for the desktop app and core (outside ee/); OpenWork EE License on the ee/ Den control plane (source-available) (source)Elastic License 2.0 — source-available, not OSI-open (source)
Form factorDesktop app for macOS, Windows, and Linux; not a CLI (source)CLI plus web/cloud; agents run on disposable cloud computers (source)
Where agents runOn your local machine, on your own files; cloud optional (source)One isolated sandbox per session, on its own branch (source)
Models50+ providers, your own keys, Ollama, or ChatGPT sign-in (source)Any provider with your own API keys, or an existing ChatGPT/Claude/Cursor subscription (source)
Team governanceOpenWork Den control plane plus the OpenWork Connect MCP gateway with per-call roles and policies (source)Change requests a human approves, deny-by-default merge, per-resource permissions, full audit trail (source)
Self-hostSelf-host the full platform, including the org control plane (source)Self-host free via kortix self-host start from Docker images (source)
PriceDesktop free; Team $10/seat/mo (first 5 seats free); Enterprise $40/user/mo billed annually (source)Self-host free; managed cloud $40/seat/mo plus usage (source)

Feature-by-feature comparison

License

License is where the two products look most alike and where the differences matter most, because neither is entirely OSI-open. OpenWork uses a directory-split license: everything outside ee/ is MIT, an OSI-approved license, while ee/ (OpenWork Den, the org control plane) is source-available under the OpenWork EE License, which requires a subscription for production use beyond the free 5-user and 30-day-eval carve-outs, and each ee/ release converts to MIT two years after publication (source). Kortix ships under the Elastic License 2.0, which is source-available but not OSI-approved: the source is public and self-hostable for free, but you may not provide the software to third parties as a hosted or managed service (source). In short, OpenWork's desktop layer is OSI-open while its control plane is source-available, and Kortix is source-available end to end under ELv2.

Execution model

Execution model is the clearest difference between the two products. OpenWork is a desktop app: agents run on your own machine against your own files, and cloud is optional (source). Kortix is a server-side system: each session gets its own cloud computer — a disposable, isolated Linux sandbox on a branch named after the session — and the OpenCode agent works there, opening a change request that you review and merge to main (source). That means OpenWork's default is "one agent on my laptop with my files," while Kortix's default is "many parallel sandboxes feeding reviewed changes into a shared git repo."

Isolation and security

Kortix is built around isolation. Every session runs on one isolated machine, secrets are encrypted at rest and injected into the sandbox at runtime, connector credentials are brokered server-side so the raw key never reaches the sandbox, and there is a full audit trail plus merge that is deny-by-default for an agent (source). Isolation is per provider: the Platinum provider runs microVMs while the default runs containers (source). OpenWork's posture is local-first by default: in desktop mode your files stay on your machine and prompts go directly to your LLM provider, while OpenWork Connect, its MCP gateway, keeps credentials in the gateway and applies roles and policies to every call (source). OpenWork does not, by default, run each agent in a server-side sandbox the way Kortix does; its hosted cloud workers are optional and run on sandboxed infrastructure (source).

Model choice

Both products are model-agnostic, which is a shared point of contrast with Claude Cowork. OpenWork works with 50+ providers, your own API keys, local models through Ollama, or ChatGPT sign-in (source). Kortix supports any provider with your own API keys and lets you reuse a ChatGPT, Claude, or Cursor subscription you already pay for (source). The practical difference is where inference happens: OpenWork sends prompts from your machine to your chosen provider by default, while Kortix runs them from the session's cloud computer — relevant if you want inference to occur inside an isolated environment rather than on your workstation.

Team permissions

Team governance is where Kortix adds the most structure. Kortix models members, groups, and roles that match your org, with per-resource permissions for both people and agents, a full audit trail, and change-request review as the default path to main (source). OpenWork handles teams through OpenWork Den and its Connect gateway: you invite teammates, create teams, control which members can use each model provider, and set roles and policies that apply to every call in every MCP client (source). Both gate access, but Kortix's gating is git-native (reviewed change requests), while OpenWork's is gateway-native (per-call role policies).

Pricing

Pricing differs in structure. OpenWork's desktop app is free, the Team plan is $10 per seat per month with the first 5 seats free and adds API access and the Extension Marketplace, and Enterprise is $40 per user per month billed annually with SSO, desktop policies, audit log, and bring-your-own inference (source). Kortix is free to self-host, and its managed cloud is $40 per seat per month plus usage (source). The two $40 figures are not apples-to-apples: OpenWork's $40 Enterprise tier is for governance features and still expects you to bring inference (or buy the $10/user OpenWork models add-on), while Kortix's $40 is for managed cloud compute. Both products can be $0 — OpenWork's desktop app and Kortix's self-hosted install — but each free option carries different caveats.

When to choose which

Choose OpenWork when you want a local-first desktop agent that runs on your own files, keeps prompts on your machine by default, and has an OSI-approved MIT core; OpenWork is the better fit for an individual or a small team of up to 5 users who do not want a server-side sandbox-and-change-request model (source). OpenWork is the wrong fit when you need governed, isolated parallel agents whose work lands through audited, human-approved change requests, or when you want your team's governance layer to be free to self-host beyond 5 users — those are Kortix's strengths (source).

Choose Kortix when you want a git-repo "company brain" where agents, skills, memory, and connectors are versioned and diffable, and where every session runs in an isolated sandbox and lands work through a reviewed change request (source). Kortix is the wrong fit when you specifically need an OSI-approved open-source license (Kortix is ELv2 source-available, not OSI), or when you want a simple point-and-click desktop app rather than a CLI-and-cloud system (source).

Verdict

Kortix vs OpenWork is less a contest of better-or-worse than a choice between two different shapes of the same category. OpenWork is a local-first desktop app with an OSI-open MIT core and a source-available team control plane; Kortix is a cloud-native system where isolated, governed agents land reviewed changes into a git repository the company owns (source) (source). If your priority is a free, local, MIT-licensed agent on your own files, OpenWork fits. If your priority is governed, isolated, parallel agents with a git-native review loop and free self-hosting, Kortix fits. Neither is fully OSI-open end to end, and both are transparent about that — the decision turns on whether you want a desktop agent on your own files or a governed, isolated company of agents.

Related reading: Kortix vs Eigent vs Kuse AI and open-source Claude Cowork alternatives.

FAQ

Is OpenWork open source?

Partially. OpenWork's desktop app and core platform — everything outside the ee/ directory — are MIT licensed, which is an OSI-approved open-source license (source). The ee/ directory, which contains OpenWork Den (the organization control plane), is under the OpenWork EE License, a source-available license: the code is public to audit, but production use requires a subscription beyond the free 5-user and 30-day-evaluation carve-outs (source).

Is Kortix open source?

Kortix is source-available, not OSI-open. Its source is public, forkable, and free to self-host, but it ships under the Elastic License 2.0, which is not an OSI-approved license and which restricts providing the software to third parties as a hosted or managed service (source). If you specifically need an OSI-approved license, OpenWork's MIT desktop core is the OSI-open option of the two (source).

Is Kortix free?

Yes, to self-host. Kortix is free when you run it on your own infrastructure — a laptop, VPS, VPC, or on-prem network — via kortix self-host start from Docker images (source). Its managed cloud is $40 per seat per month plus usage (source). Self-hosting still requires you to supply and pay for your own compute and model API keys.

Which is cheaper — Kortix or OpenWork?

It depends on what you need. Both can cost $0: OpenWork's desktop app is free, and Kortix is free to self-host (source) (source). OpenWork's Team plan at $10 per seat per month is cheaper than Kortix's managed cloud at $40 per seat per month plus usage, but they buy different things — OpenWork's $10 is for the control plane while you still bring model keys, and Kortix's $40 is for managed cloud compute (source).

Does OpenWork self-host?

Yes. OpenWork lets you self-host the full platform on your own infrastructure, including the organization control plane (source). Self-hosting the control plane is free for up to 5 users, free for development and testing, and free to evaluate for 30 days at any size, excluding Enterprise features such as SSO, analytics, desktop policies, and white-labeling (source).

Do I need to self-host Kortix?

No. Self-hosting Kortix is optional. You can sign up at kortix.com, create a project, and start sessions on Kortix's managed cloud without installing anything, at $40 per seat per month plus usage (source). Self-hosting matters when you want your agents, data, and model keys to stay on your own VPC or on-prem network, which Kortix also supports (source).