.png?width=160)
CocoIndex
Open source ETL to index fresh data for AI
FreeFramework225 impressions#1444 of its week
CocoIndex is an open-source incremental ETL framework that keeps codebases and other data continuously indexed for AI agents.
- for
- Developers building AI agents that need fresh, structured context from code, docs, and other sources.
- pricing
- open source
- license
- Apache-2.0
Key features
- Real-time incremental indexing — Reindexes only the changed delta, delivering sub-second freshness at any repo size.
- AST-based code context — Generates call graphs, hierarchies, symbol tables and semantic indexes directly from source code.
- Semantic search — Find functions, patterns or intent by meaning rather than plain text grep.
- Built-in lineage & observability — Tracks every byte back to its source and provides metrics on freshness, throughput and cost.
- Parallel task scheduling — Executes transformations in parallel by default for low-latency, low-cost processing.
- Multi-source support — Ingests code, meeting notes, APIs, databases, vector stores, files, video, audio and more.
Use cases
- Generate up-to-date code-review agents that understand call graphs
- Create knowledge-graph agents from meeting notes and decisions
- Power RAG pipelines with continuously refreshed vector indexes of repositories
CocoIndex vs alternatives
.png?width=64) CocoIndex |  GraphBit |  Metatype |  Encore |  Keel | |
|---|---|---|---|---|---|
| Best for | Incremental indexing for AI agents | Agentic AI framework | WASM-based backend building | Engineering scaling | Backend as a service |
| Pricing | Open source | Free | Free | Subscription | Subscription |
| DevHunt upvotes | 0 | 16 | 37 | 18 | 46 |
| Launched | Jan 2026 | Oct 2025 | Jun 2024 | May 2025 | Mar 2024 |
- CocoIndex vs GraphBit: GraphBit is a Rust-core, Python-first agentic AI framework, while CocoIndex focuses on incremental data indexing.
- CocoIndex vs Metatype: Metatype builds backend components with WASM and TypeScript, whereas CocoIndex indexes arbitrary data for AI agents.
- CocoIndex vs Encore: Encore helps scale engineering operations, not specifically incremental indexing of data for agents.
- CocoIndex vs Keel: Keel provides backend services and team tools, not a data-pipeline engine for fresh AI context.
CocoIndex FAQ
How does CocoIndex achieve freshness?+
It watches source files, computes a fingerprint, and reprocesses only the changed chunks, keeping the index up to date in seconds.
Do I need to write DAGs for pipelines?+
No, you write normal Python functions; the engine derives the execution graph automatically.
Can I use CocoIndex with existing vector stores?+
Yes, it can embed chunks and sink vectors to stores like pgvector, PostgreSQL, or any vector DB you configure.
Is the system production ready?+
The website states it is production-ready and can be set up in about 10 minutes.
What observability does it provide?+
It offers live lineage views, metrics on throughput and freshness, and failure management with retries and back-off.
Summarized by DevHunt from cocoindex.io · Sep 28, 2026. Details may change; check the official site.
.png?width=64)
.png?width=64)
