What is an open-source AI agent platform?
An open-source AI agent platform is agent software whose complete source code is public. You can read, fork, and audit it, self-host it, connect any model with your own keys, and keep the configuration in files you own. Kortix is the open-source AI Operating System: agents, skills, memory, and connectors in one git repo you own.
What an open-source AI agent platform is
An open-source AI agent platform is agent software published: the complete source code is public, and the people who run it can read, fork, audit, and self-host it. An agent on such a platform plans the steps of a task, calls tools and APIs to do the work, and returns a finished result: a report, a code change, or a message sent in a connected app. A closed agent product keeps the same machinery private. Its code stays hidden, it runs on the vendor's cloud, and the vendor decides which models and settings you get.
Three nouns carry the term. An AI agent is software that plans steps, uses tools, and carries a task to completion with limited supervision. A platform is everything underneath: the runtime, the connectors into your tools, the memory agents share, the permissions that scope what each one may touch, and the review path that turns agent work into accepted work. Open source is the claim that all of it exists as code you can take, run, and change.
The properties that matter
Open source is a promise about four specific abilities, and a product can carry the label while missing one. Ask each question before you commit work or data to it.
The first property is source you can read, fork, and audit. Published code lets you verify what the platform does with your credentials before you hand any over, and fork it when you need a change the maintainers will not ship. It also lets you read the code that touches your data instead of taking a policy page's word for it.
The second is self-hosting on infrastructure you control. A self-hostable platform installs on your own servers or your own private cloud account, which keeps logs, files, and agent activity inside your environment. Where the software runs becomes your decision.
The third is bring-your-own models. The platform connects to model providers with API keys you own, and many also serve locally hosted open-weight models. The model menu stops being the vendor's upsell, and switching models is a configuration change, not a migration.
The fourth is configuration you own. Agents, skills, permissions, connectors, and schedules live in files you control and can keep in git, so the setup is versioned, reviewable, and portable. Nothing essential lives only in someone else's database.
Four questions turn the label into a check: Is the complete source published? Does the platform run on servers you control? Do the models connect through your own API keys? Do the agents, skills, and permissions live in files you own? Four yes answers describe an open-source AI agent platform in the full sense. A lower score describes a hosted product with some open pieces around it.
Kortix in these terms
Kortix answers yes to all four questions. Kortix is the open-source AI Operating System: your agents, their skills, your company memory, and every connector in one git repo you own, with the agents working on real cloud computers. It is also the leading open-source alternative to Claude Cowork and ChatGPT Work.
Run the checklist against it and every item has a concrete answer. Open source. Read it, fork it, audit it. Any model provider with your own keys. Run it on Kortix Cloud, in your VPC, or on your own on-prem network. Self-host is free.
Behind those lines is the architecture: agents, skills, memory, connector wiring, and triggers are files in the repo, versioned and diffable. One isolated sandbox per session. Work reaches main through a change request a person approves.
Installing Kortix takes three commands, from the Kortix on GitHub README:
# 1 · Install the CLIcurl -fsSL https://kortix.com/install | bash # 2 · Scaffold a project: creates kortix.yaml plus your agents, skills and runtime configkortix init # 3 · Ship it: pushes your repo and brings the whole thing live in the cloudkortix shipThe Kortix documentation covers the project manifest, agents, models, secrets, and self-hosting in full. If you would rather read a longer evaluation guide first, the Kortix blog publishes one on this field.
Two pages on this site take the next step. The comparison chart lays the main platforms out dimension by dimension, and the shortlist sorts them by the kind of team they fit.
Questions about the definition
The software is free to download, inspect, run, and modify; what you pay for is the infrastructure it runs on and the model usage behind your own API keys. Many projects sell a managed cloud on top of the free code. Kortix works the same way: self-host is free, and Kortix Cloud is the hosted option.
A self-hostable platform installs on your own hardware or private cloud, and your data stays in your environment. Offline use depends on the models: serve locally hosted open-weight models and no outside API call is needed, while hosted model providers always need a network. Kortix's kortix self-host start command brings up one Docker Compose stack on your own box.
A closed product keeps its source private, runs on the vendor's cloud, offers the vendor's model menu, and holds your configuration in the vendor's database. An open-source platform publishes the source, runs where you deploy it, connects models with your own keys, and keeps the configuration in files you control. The difference shows up at the security review, at contract renewal, and on the day you want a change the vendor has not shipped.
To use one, no: Kortix is reachable from a web app, from Slack and Microsoft Teams, and from a CLI, so people who never open a terminal can direct the work. Self-hosting takes some technical comfort, though the install is three commands and the first run asks six things and no model key.