# CocoIndex

An incremental data-indexing engine that keeps context such as codebases, meeting notes, inboxes, Slack, PDFs and video continuously fresh for AI agents and LLM apps, recomputing only what changed.

CocoIndex is an incremental data-indexing engine, written in Rust and used from Python, that turns sources such as codebases, meeting notes, inboxes, Slack, PDFs and video into continuously fresh context for AI agents and LLM applications to reason over. Rather than reprocessing everything on a schedule, it recomputes only the data that has actually changed since the last run.

A user declares what should be in the target index, and CocoIndex keeps that target in sync going forward: an initial run backfills the whole source, and every later run re-embeds only the files that changed since then. It is built to run in parallel by default and is designed to scale from a single repository up to petabyte-scale collections of data.

It installs with pip, and ships a companion CocoIndex skill file that an AI coding agent can read for the concepts, APIs and patterns needed to write correct CocoIndex code.

Website: https://cocoindex.io