Lotor
Local, receipt-built agent memory that outlives the horizon
4 followers
Local, receipt-built agent memory that outlives the horizon
4 followers
Lotor is load-bearing memory for AI agents. Every session leaves a signed, hash-chained receipt on your machine, and the record is the agent: swap the model, the memory survives. Lose the record, and no model can bring it back. High-risk actions stop for a passphrase only you know. The key is never on disk. You hold the record, not the vendor. The agent can act, but it cannot sign and is witnessed by a triad ledger model.
This is the 2nd launch from Lotor. View more
Lotor v3.0
Launching today
Lotor now records every step an agent takes on your machine, down to the reasoning behind it. You can replay a run, or start at the result and walk it back to see what actually worked and what was a dead end. Then you share it as a signed cartridge and only show what you choose. It’s for anyone using agents to reverse engineer software or build on someone else’s work who needs to show how it got done.
Launch Team
Render: Get up to $100K in startup creditsThe fastest way to launch and scale your startup in the cloud
Promoted
So I’ve been watching everybody reverse engineer stuff with agents the last couple weeks, and it’s impressive. But nobody can show you how they got there. You get the result, you don’t get the method. And if you can’t replay it, you can’t really say what was good about it and what was bad.
That’s what Lotor does now. It records every step the agent takes on your own machine, with the reasoning behind it, so you can replay the whole run. Or start at the end and walk it backwards, see what actually produced the result and what was a dead end. I call that the retrograde.
Then you share it. The run goes out as a cartridge, a signed QR tied to your chain, and you pick what goes in. People get the steps and the proof, not your whole history. Identities get resolved locally first, so you’re in control of your data the whole time.
It’s built for more than one agent, too. The one doing the work and the ones checking it are all on the record, so you’re not taking any single model’s word for it.
If you want to contribute, same rule. Bring the evidence that your change fixes something and doesn’t break something else.