incredlabs
Open source (MIT) — Rust-native data infrastructure for AI

The data layer for
AI-native software

incredlabs builds open-source, high-performance databases and agent memory in Rust — so you can store, query, and reason over data at the speed AI demands.

Explore products View on GitHub

ORMDB

For backend & full-stack engineers

The database that speaks ORM natively.

Learn more →

Liath

For AI engineers & agent builders

The SQLite for AI agents.

Learn more →

What is incredlabs?

incredlabs is an open-source software company (formerly Skelf Research) building Rust-native, high-performance data infrastructure for AI. We build the databases, memory stores, and data primitives that AI-native software actually needs — fast, embeddable, and MIT-licensed.

2

public products, shipping today

MIT

licensed — no fee, no lock-in

Rust

cores — safe & predictable

Our two products — ORMDB and Liath — are developed in the open at github.com/incredlabs. Read more about incredlabs or browse the FAQ.

Two products, one data platform

Focused, composable building blocks. Use one on its own, or combine them — ORMDB for your application data, Liath for your agents' memory.

Database

ORMDB

The database that speaks ORM natively

ORMDB is an open-source, Rust-native relational database that understands your entities and relations, so a single query returns a whole object graph instead of the N+1 round-trips a traditional ORM generates.

Who it's for: Backend and full-stack engineers who already reach for Prisma, Drizzle, TypeORM, or SQLAlchemy and want the database itself to speak ORM — plus teams adding AI/vector features to an existing relational app.
  • Graph fetches in a single query
  • Native one-to-one, one-to-many, many-to-many relations
  • Safe online migrations with automatic safety grading
  • Change streams for cache invalidation and real-time
$ cargo install ormdb-server
Learn about ORMDB Product site GitHub
AI memory / Database

Liath

The SQLite for AI agents

Liath is an embedded, open-source memory store for AI agents — the SQLite for AI agents — where the agent writes sandboxed Lua to query its own memory instead of calling a fixed vector-search API.

Who it's for: AI engineers and agent builders who have outgrown a one-line semantic_search() call and want programmable, sandboxed memory they can embed without standing up a vector-database service.
  • Agents write Lua to query their own memory
  • Sandboxed execution — no file, network, or system access
  • Embedded and zero-infrastructure — no server to run
  • Rust core: fast, portable, single dependency
$ pip install liath
Learn about Liath Product site GitHub

Not sure which you need? See the product comparison.

Why we build this

Modern software is built around models, embeddings, and agents — but the data layer most teams reach for was designed for a different world. We rebuild it for the AI era.

The database fights your objects

The problem

Every app uses an ORM, and every ORM re-translates object graphs into SQL and back — the source of N+1 queries and impedance mismatch.

Our answer

ORMDB moves graph awareness into the database. Declare your model and fetch whole object graphs in one round-trip.

See how ORMDB works →

Agent memory is a fixed API

The problem

Most agent memory exposes a single search(query, k) call. Real tasks need to filter by recency, re-rank, and cross-reference — logic a fixed API cannot express.

Our answer

Liath gives the agent a sandboxed Lua runtime to write its own retrieval, ranking, and filtering — embedded, with zero infrastructure.

See how Liath works →

AI features bolted on the side

The problem

Vector search, embeddings, and agent memory usually live in separate services you sync, operate, and pay for on top of your primary store.

Our answer

We treat these as first-class primitives. Semantic search sits next to relational data; memory embeds directly into your app.

Why we build in Rust →

Capabilities across the platform

The primitives modern AI applications actually need — first-class, not bolted on.

Graph fetches, no N+1

Declare your model and pull entire object graphs in a single round-trip. The join planning lives in the database, not your app.

Programmable agent memory

Agents write sandboxed Lua to filter, rank, and combine their own memory — retrieval as code the model can adapt on the fly.

Vector, geo & full-text search

Semantic search sits next to relational data instead of a separate system — the AI primitives are built into the engine.

Safe online migrations

Every migration gets an automatic safety grade before you run it, so schema changes ship without surprise downtime.

Change streams (CDC)

Subscribe to entity and relation changes with version tokens — real-time features and cache invalidation solved at the source.

Sandboxed by design

Liath runs model-generated code with no file, network, or system access — a hard security boundary around every query.

Embedded, zero-infra

A KV store, vector search, and embeddings in a single Rust dependency — no server to run, install as pip or a crate.

Rust all the way down

Memory safety without a garbage collector and predictable performance, from embedded libraries to distributed servers.

See it in code

Both products get out of your way. Keep the code you know — ORMDB resolves the graph, Liath runs the retrieval.

ORMDB — Fetch a whole object graph in one round-trip 
// Your ORM code — but ORMDB resolves it server-side
const user = await db.user.get(id, {
  posts: { comments: true, tags: true },
  profile: true,
});
// → 1 query. No N+1. No joins to reconstruct.
Read the ORMDB guide →
Liath — The agent writes Lua to query its own memory 
-- Model-generated retrieval, run in a hard sandbox
local hits = semantic_search("mem", query, 20)
local recent = filter(hits, function(r)
  return r.age_days < 7 and r.importance > 0.8
end)
return json.encode(top(recent, 5))
Read the Liath guide →

Frequently asked questions

What is incredlabs?
incredlabs is an open-source software company building Rust-native, high-performance data infrastructure for AI applications — including ORMDB (a database that speaks ORM natively) and Liath (the SQLite for AI agents). It was formerly known as Skelf Research.
What products does incredlabs make?
incredlabs makes two public products: ORMDB, a relational database that fetches whole object graphs in one round-trip, and Liath, programmable memory for AI agents queried with sandboxed Lua. More data-platform components are in active development.
Are incredlabs products open source?
Yes. incredlabs products are open source under the MIT License and developed in the open at github.com/incredlabs. There is no license fee.
What is the difference between ORMDB and Liath?
ORMDB is a relational database that speaks ORM natively — you declare your model graph and fetch object graphs in one round-trip. Liath is programmable memory for AI agents — the SQLite for AI agents — where agents write sandboxed Lua to query their own memory. Use ORMDB for your application data and Liath for per-agent memory.
How do I get started with incredlabs?
Install ORMDB with `cargo install ormdb-server` or Liath with `pip install liath`, then read the docs at docs.incredlabs.com. Everything is open source on GitHub at github.com/incredlabs.

See the full incredlabs FAQ for more on licensing, languages, and how the products fit together.

From the blog

Notes on building data infrastructure for the AI era.

Programmable memory for AI agents: why fixed vector APIs fall short

Most agent memory is a fixed similarity-search call. Real tasks need custom retrieval — so Liath lets agents write sandboxed Lua to query their memory.

Read post →

Introducing incredlabs: the data layer for AI-native software

Why we're building Rust-native databases and agent memory for the AI era — and what ORMDB and Liath are for.

Read post →

Start building on incredlabs

Everything is open source (MIT). Install a product, read the docs, star the repos, and ship AI-native software today.

Read the docs Explore products

Building something on our stack?

Questions, partnership ideas, or feedback on ORMDB or Liath — we'd love to hear from you.

Get in touch