Is Kortix an agent framework?
No. Kortix is the open-source AI Operating System, a layer above a framework. A framework is a library you build with; Kortix runs and governs the agents, the skills, the memory and the connectors in one repo.
SpecificationOpen Source AI Agent FrameworkRev 2026-10-14
Open source agent framework alternatives, compared
An open-source agent framework is a library your engineers build with. Kortix is the open-source AI Operating System that runs and governs a company’s agents from one git repo. Here is how the frameworks most teams shortlist compare with the system that owns the whole job.
Open source agent framework alternatives at a glance
A framework is a library you build with; Kortix is the open-source system that runs and governs the agents.
| Criteria | What it is | Open source | What you own | Runs the agents or defines them | Self-host | Model choice | Human review of changes |
|---|---|---|---|---|---|---|---|
| Kortix | The open-source AI Operating System that runs agents | Yes. Read it, fork it, audit it | The whole system in one git repo | Runs them on its own cloud computers | Kortix Cloud, your VPC, or on-prem | Any model, your own keys | A change request a person approves |
| LangGraph | A low-level library for stateful agents | Yes | The runtime code you write | Defines them; you deploy and operate | On your own infrastructure | Whichever model your code calls | Human-in-the-loop you wire in |
| CrewAI | A Python framework for agents, crews and flows | Yes | The Python app you build | Defines them; you run the app | On your own infrastructure | Whichever model your agents call | A human review step you build in |
| AutoGen | A multi-agent framework in maintenance mode | Yes | The Python app you build | Defines them; you run the app | On your own infrastructure | Whichever model your code calls | Humans can work alongside the agents |
| OpenHands | A self-hosted control center for coding agents | Yes | The stack and your host directories | Runs coding agents; you operate the host | Local, Docker, VM, or cloud | Any model your backend calls | You review conversations in the app |
Project details taken from each project's own repository, checked October 2026.
The whole company in one manifest
# yaml-language-server: $schema=https://kortix.com/schema/kortix.v2.schema.jsonkortix_version: 2default_agent: kortixagents: kortix: file: agents/kortix.md connectors: all skills: all kortix_permissions: allSource: https://kortix.com/docs/project/manifest
A project is a git repo with kortix.yaml at its root; agents, skills and connectors are files in it.
How to choose between an open source agent framework and a system you run
The projects on this page are libraries. You add one to an application, write the orchestration, and host the result. Kortix is a different layer: the open-source AI Operating System that runs and governs a company's agents. The distinction decides which one you should pick.
A framework is a library your team builds with
LangGraph gives you low-level infrastructure for long-running, stateful agents, and its own repository says it "deliberately does not abstract prompts or architecture". CrewAI gives you role-based crews and event-driven Flows in Python. AutoGen pioneered multi-agent conversation patterns and is now in maintenance mode, with new work pointed at Microsoft Agent Framework. OpenHands is a self-hosted control center for coding agents that can drive several agent backends. In each case the prompts, the run environment and the review step are yours to build.
That is the right choice when the deliverable is an application. If your product is the agent runtime itself, you want the primitives the framework exposes. Two head-to-head pages cover the frameworks teams ask about most: Kortix vs LangGraph and Kortix vs CrewAI.
A system you run owns the work end to end
Kortix puts the agents, their skills, company memory and every connector in one git repo you own. Each session runs on its own isolated sandbox, on its own branch, and session work reaches main through a change request a person approves. Merge is default-deny for agents. There is no orchestration library to wire up, because running and governing the agents is the product.
Pick this layer when the deliverable is finished work the company must review. For the step-by-step path, read how to run an agent framework.
What you own and operate in each case
With a framework you operate an application: you own the code, you run the process, and you keep the host alive. With Kortix you operate a repo and a control plane: agents, skills and memory are files, the model is a configuration choice with any model provider and your own keys, and it runs on Kortix Cloud, in your VPC, or your own on-prem network. Source you can read, fork and audit is the point either way; Kortix's own is at Kortix on GitHub.
For the category question behind the choice, read framework or platform. The Kortix blog also walks through how to choose an open-source AI agent platform.
Choosing an open source agent framework
Which open source agent framework should I use?
Choose by deliverable. Embed LangGraph, CrewAI, AutoGen or OpenHands when the product is an application your team builds and hosts. Choose Kortix when the product is finished work a company runs, reviews and owns.
How does Kortix compare with LangGraph and CrewAI?
LangGraph and CrewAI are frameworks you build with; Kortix is the system that runs and governs the agents. Kortix vs LangGraph and Kortix vs CrewAI each cover one head to head.
Can I self-host these?
Each framework is open source and you run it yourself. Kortix is open source too, and self-host is free, with Kortix Cloud as the managed option.
Run the agents on a system you own
Open source, any model provider with your own keys, self-host for free.
Open source · Any model, your keys · Self-host, VPC, or on-prem