Loading...
Agent Infrastructure · Startup idea
Vector databases solved retrieval. Nobody solved memory — the layer above retrieval that knows what the agent already learned, forgot, and should re-check. That's the next category.
Why now
Every agent in production runs into the same problem: the context window fills up, the agent loses track, the user repeats themselves. The fix is a separate memory tier — write-on-significant-events, decay-by-recency, query-on-need.
The idea you could build today
A managed service that exposes three tools: `memory.write(event, importance)`, `memory.recall(query)`, `memory.forget(predicate)`. Backed by a graph + embedding store. MCP-first so any agent host can plug in.
Build stack
The three repos already trying
AI-Powered Photos App for the Decentralized Web. We are on a mission to protect your freedom and privacy.
Framework migration
+109%
14-day velocity Δ
100 contributors
Engineering hiring burst
+55%
14-day velocity Δ
97 contributors
Framework migration
+35%
14-day velocity Δ
60 contributors
Matched against the current-period startup signal panel (ai-ml, data-infrastructure). Rankings shift weekly as the underlying GitHub activity moves. Read the methodology.
The seed-round pattern hiding in the trendline
Memory-layer OSS projects (Mem0-class, MemGPT-class) with sudden contributor surges around the "importance scoring" or "decay" modules are the seed-round leads.
RAG is read-only. Memory is read-write with importance scoring and decay. Different write path, different query semantics, different product.
Use the signal, not just the idea
The repos above re-rank automatically as commit velocity, contributor growth, and new-repo creation move. Want the data feed for this idea wired into your own stack? The MCP server exposes every signal as a tool any agent host can query.
Updated 2026-05-18. The framing is editorial; the “three repos already trying” slot is generated from the live signal panel. Anonymity rule: we name public GitHub orgs, never individual founders or stealth teams.