Observed arrival · 2026-10-30
FleetRing: one SQLite database per agent, on machines you own
An open-source Rust layer for placing, replicating, and failing over small SQLite databases across machines you control.
- For
- Teams operating fleets of agents, tenants, or devices
- Worth noticing
- Its reported 9.5 ms median quorum-write acknowledgment comes from three laptop nodes over loopback; the page warns real networks add latency.
Field notes
Placement is computed from each database name using weighted rendezvous hashing, with a vnode-ring option for very large fleets; region prefixes can pin data and replicas to a cell. The page describes writes streaming through the SQLite WAL and being acknowledged at quorum, with leases, fencing, promotion, and snapshot-based rebalancing in the failure and recovery flow. The current interface is a CLI; an MCP server exposing memory.write and memory.search appears in the roadmap, not among the listed shipped pieces.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue