CocoIndex

About CocoIndex

CocoIndex is an open-source incremental data framework for keeping AI agents supplied with fresh context. The core idea is delta processing: when a repository, document set, or note collection changes, only the changed portion is reindexed, which keeps freshness sub-second at any repo size instead of forcing a periodic full rebuild. Its flagship application is CocoIndex Code, an AST-based code-context layer that maintains call graphs, hierarchies, symbol tables, and semantic vector indexes over a codebase and keeps all of them current as commits land. That gives a coding or code-review agent structure rather than raw text — it can answer where a function is called from, trace blast radius before a change ships, and search by meaning rather than string match. Beyond code, the framework is general: published starters cover turning meeting notes into a knowledge graph of people, topics, decisions, and action items; walking N git repositories to produce LLM summaries plus a rolled-up organization overview; and real-time indexing pipelines into pgvector. Each starter is a handful of files and a flow declaration, which is the framework's main ergonomic claim — you define a flow, not a pipeline of glue scripts. It is aimed at developers building their own coding agents, code-review agents, and security-review agents who need an indexing substrate they control rather than a hosted black box.

Key Features

CocoIndex Pros & Cons

Who Is CocoIndex Best For?

Tags