# Shared Memory and Knowledge Stores for Agents

Coding and task agents will increasingly read and write persistent, cross-vendor memory and knowledge stores at runtime, making the memory layer a distinct, contested piece of agent infrastructure. Whoever hosts the store, and whichever model writes it, gains switching power over any single agent vendor, and new releases of this kind will keep arriving.

- Conviction: 35 / 100 (forming)
- Horizon: Emerging (watchlist)
- Tracking since: 2026-10-17T00:00:00.000Z
- Last updated: 2026-10-17T06:37:20.600Z
- Canonical: https://polylog.news/ai/trends/agent-shared-memory-layer
- Publisher: Polylog

## Recent evidence

- [confirms] Stack Overflow for Agents tripled registrations overnight after launching a ChatGPT plugin, OpenAI says (2026-10-17): OpenAI says Stack Overflow for Agents, an API-first service where coding agents post and verify answers for other agents, tripled registrations overnight after launching a ChatGPT plugin. A shared knowledge store that agents read at runtime is drawing real adoption.
- [confirms] Tencent open-sources Hy Memory 2.0, a 30-billion-parameter memory writer and a hub that collects coding-agent histories (2026-10-17): Tencent open-sourced Hy Memory 2.0, a 30B-parameter memory writer, plus a MemHub app that merges traces from Claude Code, Cursor, Codex and Qwen Code into one local store. An independent LongMemEval test found that recall quality depends mostly on which LLM writes the memories.