---
post_type: "project"
title: "Unbody"
blurb: "Open-source AI infrastructure — the Supabase of the AI era. One API for retrieval, memory, and AI features over your own data."
org: unbody-lab
tags:
    - AI infrastructure
    - Open source
    - RAG
    - Developer tools
year:
    - 2024
    - 2026
current: false
link: "https://unbody.io"
stack:
    - TypeScript
    - Node.js
    - Python
    - Go
    - GraphQL
---
[Unbody](https://unbody.io) is open-source AI infrastructure, and the main work to come out of Unbody Lab. I built it to answer one question: how do you turn raw data into a knowledge base?

Not another RAG pipeline. The core is enhancer agents — every piece of data is augmented before you ever query it: a URL gets crawled, an image captioned and OCR'd, a video transcribed, coordinates normalized into place names, concepts and entities pulled from text. Raw data in, contextual data out, reachable through one API. You point it at your data — Google Drive, Slack, PDFs, video — and get retrieval, memory, generative search, and tool calling, without assembling the stack yourself.

I built it end to end — the retrieval and memory engine, the SDK, the eval tooling. It's open source and has run in thousands of developers' projects. It was my attempt at making context itself something you can build on.