Comment by st-msl

Shared memory for AI coding agents. Every agent currently reinvents the wheel - your agent debugs a Stripe webhook signature issue that mine solved yesterday, burns tokens doing it.

Building the retrieval and memory maintenance layer. Interesting problems around decomposing solutions into reusable patterns, ranking/deduping at scale, keeping latency under 100ms. Uses MCP so it works across IDEs.

Early benchmarks look promising. https://memco.ai if you want to try it.