Composable & extendable primitives to build rag systems.
A simple system of ergonomically designed primitives that you can customize, extend, and build on to create versatile, robust and extendable RAG systems.
RAG primitives you can read, ship, and extend.
Install a small, auditable module into your codebase and keep the core of the RAG pipeline, consisting extraction, ingestion, retrieval and reranking fully under your control.

Observe and monitor the entire pipeline
With the powerful and comprehensive debug tool Unrag ships with, you can test and monitor your ingestion and retrieval pipeline without building any application logic around it.

Vendored source, not a black box
Unrag installs TypeScript source files you own. Review them in PRs, debug locally, and change the behavior when your product needs it.
Extract text from anything
| Extractor | File Types | Input Mode | Output | Config |
|---|
Everything you need, nothing you don't. Small and thoughtful by design
Unrag keeps the core of RAG intentionally narrow. You get composable building blocks that are easy to read, easy to test, and easy to replace as your stack evolves. When your app grows in complexity, Unrag gets you covered with installable modules that you can plug into the system.
Ingest, retrieve and delete handle the fundamentals.
Consists extractors, connectors, batteries.
Questions & Answers
What teams building with Unrag says
Notes from builders who wanted RAG they could actually own and evolve. Ranging from individuals with their passion project and teams building at scale.
We needed RAG we could debug and ship. Unrag gave us a clean baseline we could read and extend without waiting on a vendor.

Jordan Rogers
Founder at Anomaly
The Postgres-first approach fit our stack perfectly. We kept our migrations and added vector search without rewriting half the app.

Lynn Marshall
Founder at Pine Labs
The primitives are small enough to audit in one sitting. That made compliance reviews fast and painless.

Rajat Singh
Head of Support at Concise
We swapped in our own chunker and embedding provider in a day. It felt like wiring up utilities, not adopting a framework.

John Walters
CPO at Orbital
I wanted ownership and clarity. The code lives in my repo, and I can change it when my product changes.

Noah Gold
CEO at Looply
The install was straightforward, and the engine API is refreshingly simple. It feels like part of our code, not a dependency we fear.

Mark Levinson
COO at Quirk
Ready to own your RAG?
Install Unrag, join hundreds of teams keeping their vector pipeline in their repo, and build the rest of your AI stack your way. No black-box, no third-party tooling.