Atomic Strata

Docs · Read the manifesto

Why Atomic Strata

AI is becoming stateful. Its memory should be infrastructure, not a hosted black box.

Configurable. Compose models, embeddings, storage, and policy around your stack.
Durable. Persist across sessions, agents, and teams.
Inspectable. Read every retrieval. Version every claim.
Governed. Scopes, audit trails, and correction workflows. First-class, not bolted on.

Get up and running, each a single page in the docs:

Core Quickstart. Run AtomicMemory Core locally from the published Docker image, then ingest and search over HTTP.
SDK Quickstart. Install @atomicmemory/sdk, point MemoryClient at Core, ingest and search typed.
CLI. Drive Core from the terminal — ingest, search, and inspect memory without writing code.
Integrations. Wire AtomicMemory into agents, frameworks, and editors.

Repo	What it is
atomicmemory	Monorepo for the Core HTTP memory engine (Postgres + pgvector, Docker) and the typed TypeScript SDK with a provider abstraction. See the package matrix.
atomicmemory-python	Python SDK for Python-native AI systems.
llm-wiki-compiler	The knowledge compiler. Raw sources in, interlinked wiki out.

We are in Developer Preview, so the surface is moving. The most useful contributions right now are issues and discussions, not pull requests.

Found a bug or rough edge? Open an issue on the relevant repo.
Have a question, an idea, or want to share what you are building? Start a discussion.
Want to chat with the team and other early users? Join us on Discord.
Follow shipping updates and design notes on @atomicstrata.

For commercial use, design partnerships, or anything sensitive, email support.

_{Apache 2.0. Built in the United States.}