Launched this week
Memora-Swarm
The TLS for autonomous agent state.
15 followers
The TLS for autonomous agent state.
15 followers
The AI giants run agent swarms on last-write-wins databases like Redis. Result? Silent state forks and swarm-wide memory poisoning (OWASP ASI06). You can't patch this with middleware. Memora replaces fragile DBs with a 3-layer mathematical engine: L1 CRDTs guarantee fork-free convergence. L3 ACFA uses Q16.16 fixed-point math to auto-ban equivocators without a central coordinator. L2 Epistemic logic detects when agents agree on a lie and forces human escalation.
Finally tried Memora-Swarm on a small agent test and the L2 epistemic layer catching a coordinated hallucination mid-run genuinely surprised me. Solid pick if you're done babysitting Redis state.
Finally something that actually tackles the OWASP ASI06 problem instead of pretending middleware can fix it. The fixed-point auto-ban layer is clever, feels like it would have saved us weeks of debugging on our last multi-agent build.