🔥 zilliztech/memsearch单周暴涨220星 — 跨Agent记忆是新赛道
GitHub Trending验证：zilliztech/memsearch 实现跨Agent统一记忆，一周获得 220星。OpenClaw/Claude Code/Codex等多个Agent框架都需要记忆持久化。
你用Claude Code开发了一个项目：
Session 1: 讨论架构，确定了Redis缓存方案
Session 2: Claude忘记所有讨论，重新提出PostgreSQL方案
Session 3: 又一次从头解释为什么用Redis
Session 4: 怎么又忘了？！
真实问题：
- 每次session重新开始，没有历史记忆
- 多Agent协作无法共享上下文
- Claude Code + Codex + Cursor 各自独立记忆
- 重要决策无法跨session持久化
Agent Memory 是一个零依赖的统一记忆层：
┌─────────────────────────────────────────────┐
│ 🧠 Agent Memory — Unified Context │
├─────────────────────────────────────────────┤
│ Sessions tracked: 23 │
│ Decisions persisted: 12 │
│ │
│ Recent decisions: │
│ ├─ Apr 30: Use Redis caching (TTL 15min) │
│ ├─ Apr 28: API rate limit 70% threshold │
│ ├─ Apr 25: Cost guard daily budget $10 │
│ │
│ Cross-agent sync: ✅ Claude Code ✅ Codex │
│ Last sync: 3 minutes ago │
└─────────────────────────────────────────────┘
# Option 1: Clone directly
git clone https://github.com/aptratcn/skill-agent-memory.git ~/.agent-skills/
# Option 2: npx install
npx skills@latest add aptratcn/skill-agent-memory
# Option 3: Copy SKILL.md
curl -o SKILL.md https://raw.githubusercontent.com/aptratcn/skill-agent-memory/main/SKILL.md<!-- Add to CLAUDE.md -->
Session protocol:
1. Start: /memory-load to restore context
2. During: Auto-capture important decisions
3. End: /memory-sync to persist| Feature | What It Does | Why It Matters |
|---|---|---|
| Decision persistence | Save architecture choices, conventions | Never re-explain decisions |
| Cross-agent sync | Share memory between Claude/Codex/Cursor | One memory, multiple agents |
| Semantic search | Find relevant past discussions | Instant context retrieval |
| Auto-capture | Detect and persist key decisions | Zero manual effort |
| Markdown storage | Human-readable, git-trackable | Version control your memory |
.agent-memory/
├── decisions/
│ ├── 2026-04-30-redis-caching.md
│ ├── 2026-04-28-rate-limits.md
│ └── 2026-04-25-cost-budget.md
├── sessions/
│ ├── 2026-04-30-session-1.md
│ └── 2026-04-30-session-2.md
└── index.json ← Semantic search index
/memory-load [topic] → Restore context about [topic]
/memory-save [decision] → Persist important decision
/memory-search [query] → Semantic search across memory
/memory-sync → Sync to all connected agents
/memory-history → Show recent memory activity
Before Agent Memory:
❌ 每session重新解释架构 (平均15分钟)
❌ 多Agent协作上下文断裂
❌ 重要决策口头传达后遗忘
After Agent Memory:
✅ /memory-load → 3秒恢复完整上下文
✅ Claude + Codex 共享同一决策历史
✅ Git-trackable决策记录，可回溯
- Session Context Bridge — 跨session上下文传递
- Token Optimizer — Token压缩优化
MIT License — Copy freely, modify as needed.
⭐ 如果这个skill帮你省了重复解释的时间，给个Star！
关联trending:
memsearch(+220星/周) 验证跨Agent记忆是新赛道