← All frameworks
47k ★WorkflowModerate learning curve

LlamaIndex

LlamaIndex · Python / TypeScript · MIT

LlamaIndex started as a retrieval library and grew an agent layer, so it shines when the hard part of your agent is connecting to and querying your own data.

Best for

Agents whose main job is reading your documents, databases, and knowledge bases.

Strengths

  • +Strongest retrieval and indexing primitives in the field
  • +Event-driven Workflows API for agent control
  • +Hundreds of data loaders and connectors

Watch out for

The retrieval surface is huge; you can over-engineer a pipeline before writing any agent logic.

Spec sheet

Vendor
LlamaIndex
Languages
Python, TypeScript
License
MIT
Architecture
Workflow
Model support
Any provider
Multi-agent
Yes
State persistence
Yes
Human-in-the-loop
Yes
Streaming
Yes
Observability
Built-in + integrations
Memory
Built-in
First released
2023

Compare with

This page reflects LlamaIndex's public documentation as of May 2026 and is maintained by Choppy Toast, which is not affiliated with LlamaIndex. GitHub star counts are approximate. Always confirm details in the official docs before you commit.