Migrate from OpenWork to an open-source system your team owns
Switching a team from OpenWork to Kortix is a rebuild of four things in one open-source home: skills, MCP connections, memory, and model keys. This is the order to move them in, with the command or file the Kortix docs give for each one.
The move, in the order the docs run it
Six steps to rebuild a team's setup. Each command and file is quoted from a Kortix docs page linked on the panel.
$ kortix init my-agent && cd my-agent$ kortix ship$ curl -fsSL https://kortix.com/install | bash$ kortix self-host init --domain kortix.example.com --admin-email you@example.com$ kortix self-host startskills/ my-skill/ SKILL.mdconnectors: - slug: docs-search provider: mcp url: https://mcp.example.com/mcpmemory/ MEMORY.md$ kortix providers set anthropic$ kortix providers login openaiWhat carries over from OpenWork, and what does not
Skills move cleanly. OpenWork keeps a skill as a SKILL.md file, and Kortix reads a skill from skills/<name>/SKILL.md, so a skill you wrote drops into the new project and loads the same way. Its reference files sit beside the SKILL.md, exactly as they did before.
Your local files do not move at all. OpenWork keeps them on the machine that opened them, and Kortix asks for no duplicate copy either: a Kortix project holds the agent config and state, and when a task needs a codebase the agent clones that repository into its own session. The files the team already has stay where they are.
MCP servers have to be reconnected. OpenWork states that credentials do not move between products, and Kortix holds each credential server-side in a connector, so every service is authorized again inside the project. Connector credentials are brokered server-side and never enter the machine.
Memory is the part a team rebuilds by hand. OpenWork's cross-chat memory is saved session history inside the app, not a separate long-term store, so there is nothing to export. Kortix keeps memory as plain files under memory/, with MEMORY.md as the index, and a person writes down what the team knows so the next session starts with it.
The switching decision: what a team gains
The move pays off when work has outgrown one person's laptop and has to be shared, reviewed, and run with nobody at the keyboard. Kortix is the open-source AI Operating System: agents, skills, company memory, and connectors live in one git repo you own, so the setup is versioned and diffable instead of scattered across desktops. A team reads a change to an agent the same way it reads a code change.
Session work reaches main through a change request, and merge is default-deny for agents, so a person sees the diff before anything lands. Each session gets its own isolated sandbox and its own branch, which means an agent can install, run, and break things without touching anyone else's work. When an agent has to reach the tools a business already runs on, connectors cover 3,000+ apps in a click, plus MCP, OpenAPI, Postman, GraphQL and raw HTTP.
What the team has to set up
A migration is a small setup project, not a button. You need a git project, a model key or a subscription the team already pays for, and the connector authorizations each service requires. Self-hosting adds a sandbox provider key, because agent sessions run off the box on a separate sandbox provider rather than on the machine running the stack.
The tooling covers the rest. The CLI scaffolds the project, validates the manifest, and ships it; the first run creates the cloud project, and after that plain git works. A team can start on Kortix Cloud and self-host later without changing the repo, because a self-hosted instance runs the same stack from the same images.
The order to move in
Move one workflow at a time and keep the old tool until the new one passes. Scaffold the project and connect a model first, then bring the skills a team relies on, one at a time, checking that each one triggers on a real task. Re-authorize the MCP servers next, rebuild memory from what the team writes down, and recreate the schedules last, running each one by hand before it repeats.
The reason to move off a desktop agent is that sharing is the hard part. OpenWork's desktop server listens only on loopback, and its own docs say the desktop app offers no shared or remote workspace, so a team that needs one system has to run a control plane as well. Kortix starts from the server side: one project, one repo, and a review gate, which is the shape a small team grows into. The Kortix team's own Kortix vs OpenWork note walks the same decision from their side.
For the install steps themselves, read the self-hosting guide. For the tool-for-tool view that decides whether to move at all, read the OpenWork vs Kortix comparison.
Questions a team asks before it switches
Short answers, each sourced to the page it comes from.
Do our OpenWork chats and memory move to Kortix?
No. OpenWork's cross-chat memory is saved session history inside the app, not a portable store, so there is nothing to export. Kortix memory is files under
memory/, so the team writes down what it learned and commits it.Can we keep our OpenWork skills?
Yes. OpenWork and Kortix both read a skill from a
SKILL.mdfile, so a skill you wrote goes intoskills/<name>/SKILL.mdin the new project, with its reference files beside it.Do we have to self-host Kortix?
No. Run it on Kortix Cloud, in your VPC, or on your own on-prem network, and self-host is free. The repo is the same either way, so a team can start managed and move onto its own box later. The self-hosting guide lists what the instance puts on disk.
What does the move cost?
The tooling and the migration are free. Free is $0 with 200 credits per month and 1 project, and the team plan is $40 per seat per month with 2,500 credits per month per seat, pooled. Compare against OpenWork's plans before you commit.
Move your team onto one repo you own
Start on the free plan, then grow into it one workflow at a time.
Open source · Any model, your keys · Self-host, VPC, or on-prem