Run Lodestone yourself.
Everything Lodestone does can run on hardware you choose: a laptop for capture and distillation, plus an optional always-on server so claude.ai and your other machines can reach the same memory. During early access we give self-hosters the software and the setup guide on request.
What you'll need
| Python 3.11+ | For the lodestone CLI and server. The core uses only the standard library. |
|---|---|
| A private git repo | Holds your memory as plain files. GitHub works; so does any git host. |
| A model for distillation | Your Claude subscription through Claude Code (default), the Claude API, or any OpenAI-compatible endpoint, local models included. |
| A host (optional) | Any container host that gives you an HTTPS address and a persistent volume, for remote access. A Dockerfile and a Railway config are included. |
What setting up involves
Install the CLI and create your memory repo
One command lays out
archive/,memory/,digests/andconfig/in a private repo you own.Capture Claude Code sessions
A Claude Code plugin adds hooks that copy each transcript in milliseconds and never fail your session. Redaction, archiving and distillation run in the background when the session ends.
Map projects to namespaces
Namespaces keep each project's memory apart. Sessions in a folder go to that project; facts about you go to
me.Put it online (optional)
Deploy the container behind HTTPS with a persistent volume, then connect claude.ai, Codex, VS Code and other MCP clients with a scoped token per device. Never expose it over plain HTTP: the server refuses writes when storage or backups are at risk.
Check everything
lodestone doctorchecks the data repo, push access, hooks, the distillation model and search, and tells you how to fix anything that's wrong.
Request access
Tell us a little about your setup and we'll send you access to the software and the full setup guide.
Rather not run a server?
Lodestone Cloud is the same Lodestone, hosted and looked after for you.