LlamaDeploy is an open-source framework from the LlamaIndex team with 2,000+ stars on GitHub for taking agentic workflows out of your notebook and into production as deployed, callable services.
Building an agent workflow with LlamaIndex is one thing; running it reliably for other people is another. LlamaDeploy addresses the second half of the problem. It lets you package agentic and multi-agent workflows as services with defined interfaces, deploy them, and call them from client code, so the workflow you prototyped becomes an API other systems can depend on. That separation — workflow logic on one side, serving, session handling and invocation on the other — is what separates a demo from a product, and it is the gap this framework fills within the LlamaIndex ecosystem. It is MIT licensed, maintained by the run-llama organisation, and under active development with updates through 2026. Teams already invested in LlamaIndex for retrieval and agents will find it the natural deployment companion; teams on other frameworks should note it is designed around LlamaIndex workflows first.
LlamaDeploy is free and open source under the MIT license for self-hosting. LlamaIndex also offers the hosted LlamaCloud platform with paid tiers for teams who prefer managed infrastructure, making the overall ecosystem freemium.
Developers who have built something real with LlamaIndex agents or workflows and now need to serve it to users or other systems, without hand-rolling the deployment, session and API layer themselves.