Framework or platform: where an open source AI Operating System fits

A framework is a library. You add it to an application, write the orchestration, and decide how each agent runs. LangGraph, CrewAI, AutoGen and OpenHands all sit here. They give you primitives, and the prompts, the run environment and the review wiring stay yours. Choose a framework when the product is the application and engineering wants control over every step. The open source agent framework alternatives page compares the four in detail.

A platform runs and governs the agents. Kortix is the open-source AI Operating System: agents, skills, company memory and every connector live in one git repo you own, and each session runs on its own isolated sandbox, on its own branch. You configure the agents instead of orchestrating them, and the platform carries the runtime, the review and the audit. It runs on Kortix Cloud, in your VPC, or on your own on-prem network.

A framework hands the work back to your code. It does not give each run its own isolated machine, a branch, a change request or an audit trail unless your team builds those. Durable state, shared memory and human review are primitives you assemble. The platform does not enforce them for you.

Work a company must review needs a place to review it. Kortix gives every session its own cloud computer and branch, then routes the result through a change request a person approves before it reaches main. That is the difference between an agent your engineers run and a workforce the company governs.

No. Kortix does not ask you to import a library or write orchestration. It runs the agents, and you own the configuration as files. If you want the primitives to build your own runtime, a framework is the right tool. If you want the agents run, governed and reviewed, Kortix is the layer above it.