It keeps my three machines in sync, so every agent I run has the same memory and secrets, whichever machine it's on.
Mac mini
Always on.
The PC
It has the GPU, so it runs the local models.
My laptop
Wherever I am.
Every agent on all three gets the same memory, secrets and registry.
The code is private.
The code is private, so this is what I can show.
One brain, three machines. Synapse gives every agent on my fleet the same memory, secrets and registry, synced over a private network.
State replicates through one Go daemon that runs the same way on every device. Services like local inference or decryption get called on whichever machine owns the resource.
Anything privileged waits for a human approval, and I can give it from my phone. I built it in about a month, and my daily work runs on it.
Three things every agent shares, wherever it runs.
Every agent works from the same memory, on any of the three machines.
Every agent gets the same secrets. Anything privileged waits for an approval I can give from my phone.
Every agent sees the same registry, so a call for a service like local inference or decryption goes to whichever machine owns it.
It started as something called Loci.
I started it in March 2026, two weeks after I presented LODA at NASA JPL. It was a shared brain for my machines that synced by pulling from git. It worked, but syncing by hand was fragile.
Fixing that is why I built Synapse, and it took about a month. Loci is retired now, and I keep it around for the history.
It's the setup my own agents run on, and where I try out ideas from my research on orchestration and harness design.
There are 40 more on the full list.