Docs · Read the manifesto

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.

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.

Website · Docs · Blog · Discord · X

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