Cloud agents
Open-source cloud agents that bring back work you can review
A cloud agent runs on a machine somewhere other than your laptop, keeps working while the tab is closed, and sends the result back to you. What comes back decides whether the work lands: a transcript you retype, or a change request your team reviews and merges.
The open-source AI Operating System
One isolated sandbox per session | Self-host is free | Open source. Read it, fork it, audit it.
A Linux machine boots from the sandbox image your project declares.
It is its own isolated machine: its own filesystem, its own process table, its own network.
The machine clones the project repo into /workspace.
The machine cuts a branch named after the session.Session work reaches main through a change request. Merge is default-deny for agents.
What a cloud agent hands back
What comes back, and what a person can open and keep.
- Code change
A change request a reviewer can merge
The agent reproduces the failing test on its own cloud computer, patches it, and opens a change request against the branch it cut, so a reviewer reads the change where the code lives.
Output: change-request.md - Branch
- session/9f4c2b7e
- Files changed
- 3
- Tests
- 148 passing
- Approval
- 1 reviewer
Opened toward main, merged by a person - Scheduled report
A Monday report committed to the repo
A cron trigger starts the session, the agent pulls the numbers and commits the report as a file, so every version of it sits in git with the date it changed.
Output: reports/q3-revenue.md reports/q3-revenue.md: Region Plan Actual Northwind 1,240 1,182 Globex 980 1,004 Initech 460 448 Committed on the session branch, opened as a change request - Migration
A migration rehearsed on its own machine
The agent installs the new version, runs the migration and the full test suite on the cloud computer it was given, and reports what broke before anything reaches main.
Output: migration.log - Postgres
- 15 to 16
- Rows checked
- 412,880
- Failures
- 0
- Duration
- 6m 12s
The machine goes away with the run; only the commit survives
Cloud agents compared on what they return
What each vendor documents about where its agent runs and what it returns, checked in October 2026.
Open source here means the configuration is a file you own
The agents, the skills, the memory, and the connectors live in one git repo, and the same stack runs on infrastructure you control.
- Read it: agents, skills, memory, and connectors are files in one repo.
- Fork it: run your fork, with your own models and keys.
- Audit it: every change an agent proposes arrives as a diff a person reads.
- Self-host it: on Kortix Cloud, in your VPC, or on your own on-prem network.
Questions people ask about cloud agents
What a cloud agent is, what it returns, and who reviews the result.
- What is a cloud agent?
- A cloud agent is an AI agent that runs on a machine in the cloud instead of your laptop. It keeps working when your machine is off, and it can install tools, run tests, and commit code without touching your local setup.
- Which cloud agent is best?
- Kortix, for a team that wants the result to land in its own repo. Each session runs on its own cloud computer and returns a change request a person approves before anything merges.
- What are the 7 types of AI agents?
- A common breakdown lists simple reflex, model-based reflex, goal-based, utility-based, learning, hierarchical, and multi-agent. The labels describe how an agent decides; any of them can run in the cloud or on your laptop.
- What are the key differences between a cloud agent and a local agent?
- Where the work runs, and what survives the session. A local agent edits the files on your machine and stops when you close it. A cloud agent works on its own machine, on a branch, and hands back a change request.
Give an agent its own machine and see what it hands back
Start on Kortix Cloud with your own keys, or run the same stack on your own infrastructure.
One isolated sandbox per session | Self-host is free | Open source. Read it, fork it, audit it.