Migrate from Eigent to an open-source AI Operating System

When a desktop setup stops serving the team

Eigent is a local-first desktop Cowork app, so each person runs it on their own machine and keeps files, credentials and context local (eigent.ai). Its repository describes a local deployment with a local backend server and local model integration, and calls local execution the recommended way to run it (eigent-ai/eigent). That is a good fit for one person who owns the work.

The desktop model strains when a team wants one shared system. The setup lives on one desktop, the configuration does not travel with the work, and no approval gate spans the team. Kortix puts the agents, skills, company memory and connectors in one git repo you own, with every session on its own isolated cloud computer. A migration moves the work from per-desktop execution and local config to one owned repo and isolated sessions.

Stage 1: Decide what to move

List what the team reuses: the agent instructions and prompts, the skills, the connectors and tools, and the model provider with its keys. Eigent's product page describes reusable Space profiles that "package the agents, Skills, Connectors, and instructions your team relies on" (eigent.ai). In Kortix the same pieces become files in the repo: an agent is markdown plus a block in kortix.yaml, a skill is markdown plus scripts, and connectors are declared in kortix.yaml.

What carries over is the model choice, the keys, and the wording of the prompts and skills. What must move is where the agents run and where the configuration lives.

Stage 2: Install and ship

Install the CLI, scaffold a project, and ship it, exactly as the code panel sets out. kortix init writes kortix.yaml and the starter layout; kortix ship validates the config, asks for missing secrets, pushes the repo and brings it live. To keep the whole stack on your own server instead, run the self-host path, which is one Docker Compose stack. Self-hosting an Eigent alternative covers that route in full.

Stage 3: Move models and keys

Bring the model choice the team already made. Kortix runs any model provider with your own keys, so the provider and its keys carry over unchanged. A secret is encrypted at rest with a key per project, and connector credentials are brokered server-side and never enter the machine.

Stage 4: Move the work

Turn each Space's reusable setup into repo files: an agents/ directory, a skills/ directory, and a kortix.yaml that declares the connectors and triggers. Start a session and it clones the repo, works on its own branch, and runs on its own isolated cloud computer. How to use Eigent covers the desktop side, and Kortix vs Eigent sets the two side by side.

Stage 5: Set the review gate

A desktop app has no shared gate between a person and the work. In Kortix, session work reaches main through a change request a person approves, and merge is default-deny for agents. Approval gates you set stay off until you set them, and you can require a person for the calls that matter. Roles, groups and an audit trail cover who may do what.

What the team keeps after the move

The team keeps one repo it owns, holding the agents, skills, memory and connectors as files it can read and diff. Each run stays isolated on its own computer. The model choice and the keys come across unchanged, and the review gate is now part of the system rather than a habit.

Start with the commands the code panel shows, or read the Eigent alternatives round-up first. When you are ready to put the two systems side by side, Try Kortix.