Announcement:

Revell requires webhooks, API routes, and an agentic framework in order to function. There cannot be a “local only” version of this. The very thing that makes it superior to MemPalace or markdown files is reliant on it connecting to the internet.

If you have an agentic framework (Claude Code, OpenClaw, Hermes, LangChain, etc.) you are using the internet. The gateway by which your agent is able to exist is already online. If you use an API for your LLM layer, and I assume you do, that’s also on servers somewhere.

There is no such thing as a truly all local agent unless you’re some freak computer enthusiast who lives in the artic circle and owns his own server tower, who is also able to keep an 8b parameter model locally hosted on his SkyNet super computer, and who also wants such extreme data privacy that he doesn’t ever connect to the internet.

At that point — yes — you’ve come that far, look for a local-only memory system. If that is not your situation, you’re already exposed to the internet — at least use a memory system that works.

_______________________________________________________________

Your data is private. We do not sell, share, or allow anyone to ‘train’ on your data. I cannot see your dashboard. I cannot see your agent’s memories. There is one LLM inference layer — to help programmatic and deterministic security systems work better — who can see anything your agent/companion remembers. And that’s it. But I can’t keep defending the trustworthiness of the internet. It’s the internet guys. We lock down your data, and if you leave, we wait a few months and then delete it completely. We are obligated by law to do this. We are also obligated to delete any user data upon their written request. We follow all privacy laws. But we have to use the internet. There’s just no magical way to bootstrap a payload without the internet.