Choosing an open source AI agent framework often comes down to Kortix or LlamaIndex. LlamaIndex is a Python toolkit for building retrieval and agents over your documents. Kortix is the open-source AI Operating System: your agents, skills, memory, and connectors live in one git repo you own, and each session runs on its own cloud computer.
LlamaIndex's cells state what its own repository and docs document, and the Kortix column quotes Kortix.
| Criteria | What it is | Open source | What you own | Runs the agents or defines them | Self-host | Model choice | Human review of changes | Primary strength |
|---|---|---|---|---|---|---|---|---|
| Kortix | The open-source AI Operating System that runs agents and lands reviewed work | Yes: read it, fork it, audit it | Agents, skills, memory, and connectors in one git repo you own | Runs them: one isolated sandbox and branch per session | Self-host is free; or Kortix Cloud, your VPC, on-prem | Any model provider with your own keys, or the ChatGPT plan | Session work reaches main through a change request a person approves | Thousands of agents in parallel on one config, each isolated |
| LlamaIndex | Python data framework for RAG and agents over your documents | Yes: the framework is open source, and LlamaIndex's focus is now LlamaParse | The app you write; you run and persist its storage yourself | Defines them: you assemble the parts and run the application | Yes, a Python package you install and run on your own machines | Your preferred LLM, embedding, and vector store, via 300+ integrations | Workflows support human-in-the-loop review you wire into the app | Data and RAG: connectors, indices, and retrieval over your documents |
LlamaIndex's cells come from its own repository and framework docs, checked October 2026. Verify them before you commit.
# kortix.yamlagents: support-triage: file: agents/support-triage.md connectors: alltriggers: - slug: daily-digest type: cron cron: '0 0 9 * * 1-5'An agent and a scheduled trigger in kortix.yaml, the Kortix layer of a project you own.
LlamaIndex is an open-source Python framework for building LLM applications over your own data. It began as a data framework for retrieval-augmented generation, and it still provides the parts for that job: connectors that ingest sources such as APIs, PDFs and SQL, indices that structure the data, and a retrieval interface that returns context to a model. The framework also carries an agent layer, where agents and event-driven workflows call tools, and a RAG pipeline can be one of those tools. Its repository states that the company's current focus is LlamaParse, its document parsing platform, while the framework stays available as an open toolkit you assemble in your own application.
LlamaIndex gives you composable parts and expects you to write the application that runs them. Kortix is the leading open-source alternative to Claude Cowork and ChatGPT Work, and it runs the agents for you. In Kortix, an agent is a markdown file with a prompt and a scoped reach into tools and resources, plus a governance block in kortix.yaml. Each session gets its own isolated machine and branch, and thousands of agents can run in parallel on one config. The Kortix docs describe the project, session and change-request model.
With LlamaIndex, you own the application code you write, and you run the pipeline and its storage, which starts in memory and persists to disk when you choose. With Kortix, you own the repository: agents, skills, memory and connector configuration are files, versioned and diffable, and the whole team can read them. Both can run on your own infrastructure. LlamaIndex is a Python package you install and operate. Kortix runs on Kortix Cloud, in your VPC, or on your own on-prem network, and self-host is free.
A team stays with a library while the work is building the retrieval pipeline and the first agents. It moves to a platform when the pipeline exists and the problem becomes running many agents, governing what each one may touch, and reviewing what they change. That is where Kortix fits, with work reaching main through a change request a person approves. Kortix does not replace the retrieval and document tooling a team already runs; teams keep their data pipeline and add Kortix to run the agents and review what they change. If you are weighing other agent stacks, the Kortix vs CrewAI comparison and the open-source agent framework alternatives cover the rest of the field.
They do different jobs, so most teams do not swap one for the other. LlamaIndex is a Python framework for building retrieval and agents over your documents. Kortix is the open-source AI Operating System that runs agents, scopes what each may touch, and lands their work through review. A team can keep LlamaIndex for data and add Kortix for the agents.
LlamaIndex gives a Python library of composable parts for context augmentation, with 300+ integration packages for LLMs, embeddings and vector stores, plus document parsing through LlamaParse. It is a toolkit you build an application with. Kortix is the platform that runs the agents and lands their work. Use LlamaIndex to build and operate the retrieval layer; use Kortix to run the agents on top of it.
Yes. Run it on Kortix Cloud, in your VPC, or on your own on-prem network, and self-host is free. The install script and the kortix self-host start command bring up the full stack on your own box. Read the docs for the steps.
Yes. Kortix is open source: read it, fork it, and audit it at Kortix on GitHub. Your agents, skills, memory, and connectors are files you can read, and you can self-host the whole stack.
Start a project, run a session, and review the change request it opens.