Tencent open-sources Hy-Memory 2.0 for cross-agent shared memory
aigclink · x · 2026-10-17
Tencent has open-sourced Hy-Memory 2.0 (493 stars on GitHub), a production-grade dual-system cognitive memory engine and Python SDK for LLM agents, along with its companion app MemHub. MemHub consolidates conversation history and work experience scattered across AI tools into a unified memory system, so an agent like Codex can pick up context from prior Cursor sessions instead of starting from scratch.
The SDK ships specialized memory pipelines for three scenes:
- Chat Memory: hierarchical long-term cognitive memory — extracting cross-session facts, organizing long-term memory, and deduplicating/conflict-resolving facts;
- Working Memory: two-layer memory separating task facts from reusable experience, including the conditions under which experience applies;
- Coding Memory: version-evolution diff-chain tracking that remembers why code was changed.
Repo includes hymemory SDK, memhub app, plugins, scripts and tests under a permissive license.
Related event: Tencent Open-Sources Hy-Memory 2.0 for Shared Agent Memory(2 posts)→
More from coding & agent
- Aside v2 sneak peek: agentic browser turning into a general harness — brandon_galang · 2026-10-17
- Free library of 70+ code-based video components for agentic one-shot editing — eptwts · 2026-10-17
- greentext of an AI coder's compute doom loop: more boxes, CI bills, and no power left — peterjliu · 2026-10-17
- Superpowers beats Agent Skills in head-to-head test: same app in 49 min for $4.76 vs 90 min for $6.30 — chessbuzz · 2026-10-17
- Pydantic Monty ships TypeScript bindings to sandbox model-written Python in Node and browsers — samuelcolvin · 2026-10-17
- LangChain's deepagents adds Gemini support with 1M tokens and agentic video understanding — hwchase17 · 2026-10-17