We're open source. If actrone-memory has been useful to you, a star on GitHub means a lot to us.

about actrone

Agents should be as accountable as they are autonomous.

We build infrastructure that makes AI agents safe to run in production. We started with memory, the piece every agent needs first, and released it as two open-source libraries. The governed platform that runs agents on top of that memory is in development.

Available today

Two MIT-licensed libraries that share one memory model. Both install with one command and run without any services.
  • Python and TypeScript

    actrone-memory for Python 3.11+ and actrone-memory for Node.js 22+. One two-tier memory model, with matching result shapes in both languages.

  • Local-first

    Runs in your process with an in-memory store and a local embedder. No API key, no database and no account to start. Add Redis and Qdrant when you need durable storage.

  • Framework adapters

    16 Python adapters ship as optional extras and 11 TypeScript adapters ship as structural helpers. Check the integrations page for the framework you use.

How we work

Four commitments that shape what we ship, and what we say about it.
  • Open source, MIT licensed

    Both memory libraries are MIT licensed. Use them, fork them and ship them in commercial products.

  • Honest about limits

    The homepage lists what the libraries do not do. We publish our own recall eval with its caveats, and no head-to-head figures against other libraries until we have run them.

  • Your data stays where you put it

    The libraries send nothing to Apocalypse Technologies, the company behind Actrone. Data goes only to the stores and providers you configure, and with the defaults it never leaves your machine.

  • Governance by design

    The hosted platform is being built so every agent action clears policy before it runs and leaves an audit trail after it. That is the product we are working toward.

What comes next

The libraries are the foundation. The hosted platform builds on them.
  1. available now

    Open-source memory libraries

    Two-tier memory for Python and TypeScript, with hybrid recall, token budgets and PII-aware fact extraction.

  2. in development

    The hosted, governed platform

    Managed memory, policy enforcement before each agent action, and an audit trail for agents in production.

  3. early access

    Founding 500

    The first 500 people on the waitlist get a launch discount, priority onboarding and a direct line to the team that decides what ships first.

Build with us

Try the libraries and tell us what breaks, on the Python repository or the TypeScript repository. Or reserve early access to the hosted platform.