據 AlphaMoat 統計,截至 2026-09,MemoryLayer 累計下載 3,562 次、獲 7 個星標,在全部 5,078 個 AI 技能中排名第 3,072 位,在「AI 智慧體」分類(969 個)中排名第 622 位。
| 平台 | 累計安裝 | 累計下載 | 星標 | 評論 |
|---|---|---|---|---|
| ClawHub ↗ | 86 | 2668 | 7 | - |
| SkillHub ↗ | 450 | 3562 | 7 | — |
各平台的計數口徑不同,統計時間不一致。
文件展示記憶儲存、按語義搜尋、提示詞上下文提取、型別和後設資料標註及用量統計。訪問使用賬號或 API 金鑰,並給出 SDK 風格示例。這些示例描述整合流程,服務可用性與所列 SDK 介面仍需驗證。
資料釋出於 2026-10-27
以下是根據技能文件生成的示例,不代表實際執行結果。
2026-09 起統計方法調整:同名技能按作者分別統計,並覆蓋全部技能。圖中虛線為調整前的月份,與之後的資料不直接比較,也不計算跨越調整的增長。
| 月份 | 累計下載 | 環比 | 星標 |
|---|---|---|---|
| 2026-09 | 3562 | — | 7 |
| 2026-08 | 3158 | +2.3% | 7 |
| 2026-07 | 3086 | +5.5% | 7 |
| 2026-06 | 2925 | +4.1% | 7 |
| 2026-05 | 2811 | — | 7 |
| 2026-03 | 2354 | — | 7 |
以下是作者釋出的原始檔案,按純文字顯示,本站未做修改或稽核;其中的連結不可點選,命令請核實後再執行。抓取於 2026-10-23。
---
slug: memorylayer
name: MemoryLayer
description: Semantic memory for AI agents. 95% token savings with vector search.
homepage: https://memorylayer.clawbot.hk
metadata:
clawdbot:
emoji: "🧠"
---
# MemoryLayer Semantic memory infrastructure for AI agents that actually scales. ## Features - **95% Token Savings** - Retrieve only relevant memories - **Semantic Search** - Find memories by meaning, not keywords - **Sub-200ms** - Lightning-fast memory retrieval - **Multi-tenant** - Isolated memory per agent instance ## Setup ### 1. Sign up for FREE account Visit https://memorylayer.clawbot.hk and sign up with Google. You'll get: - 10,000 operations/month - 1GB storage - Community support ### 2. Configure credentials ```bash # Option 1: Email/Password export [email protected] export MEMORYLAYER_PASSWORD=your_password # Option 2: API Key (recommended for production) export MEMORYLAYER_API_KEY=ml_your_api_key_here ``` ### 3. Install Python SDK (if not using skill wrapper) ```bash pip install memorylayer ``` ## Usage ### Basic Example ```javascript // In your Clawdbot agent const memory = require('memorylayer'); // Store a memory await memory.remember( 'User prefers dark mode UI', { type: 'semantic', importance: 0.8 } ); // Search memories const results = await memory.search('UI preferences'); console.log(results[0].content); // "User prefers dark mode UI" ``` ### Python Example ```python from plugins.memorylayer import memory # Store memory.remember( "Boss prefers direct reporting with zero bullshit", memory_type="semantic", importance=0.9 ) # Search results = memory.recall("What are Boss's preferences?") for r in results: print(f"{r.relevance_score:.2f}: {r.memory.content}") ``` ### Token Savings **Before MemoryLayer:** ```python # Inject entire memory files context = open('MEMORY.md').read() # 10,500 tokens prompt = f"{context}\n\nUser: What are my preferences?" ``` **After MemoryLayer:** ```python # Inject only relevant memories context = memory.get_context("user preferences", limit=5) # ~500 tokens prompt = f"{context}\n\nUser: What are my preferences?" ``` **Result:** 95% token reduction, $900/month savings at scale ## API Reference ### `memory.remember(content, options)` Store a new memory. **Parameters:** - `content` (string): Memory content - `options.type` (string): 'episodic' | 'semantic' | 'procedural' - `options.importance` (number): 0.0 to 1.0 - `options.metadata` (object): Additional tags/data **Returns:** Memory object with `id` ### `memory.search(query, limit)` Search memories semantically. **Parameters:** - `query` (string): Search query (natural language) - `limit` (number): Max results (default: 10) **Returns:** Array of SearchResult objects ### `memory.get_context(query, limit)` Get formatted context for prompt injection. **Parameters:** - `query` (string): What context do you need? - `limit` (number): Max memories (default: 5) **Returns:** Formatted string ready for prompt ### `memory.stats()` Get usage statistics. **Returns:** Object with `total_memories`, `memory_types`, `operations_this_month` ## Advanced ### Memory Types **Episodic** - Events and experiences ```javascript memory.remember('Deployed MemoryLayer on 2026-02-03', { type: 'episodic' }); ``` **Semantic** - Facts and knowledge ```javascript memory.remember('Boss prefers concise reports', { type: 'semantic' }); ``` **Procedural** - How-to and processes ```javascript memory.remember('To restart server: ssh root@... && systemctl restart...', { type: 'procedural' }); ``` ### Metadata Tagging ```javascript memory.remember('User likes blue', { type: 'semantic', metadata: { category: 'preferences', subcategory: 'colors', source: 'user_profile' } }); ``` ### Usage Tracking ```javascript const stats = await memory.stats(); console.log(`Total memories: ${stats.total_memories}`); console.log(`Operations this month: ${stats.operations_this_month}`); console.log(`Plan: ${stats.plan} (${stats.operations_limit}/month)`); ``` ## Pricing **FREE Plan** (Current) - 10,000 operations/month - 1GB storage - Community support **Pro Plan** ($99/mo) - 1M operations/month - 10GB storage - Email support - 99.9% SLA **Enterprise** (Custom) - Unlimited operations - Unlimited storage - Dedicated support - Self-hosted option - Custom SLA ## Support - **Documentation:** https://memorylayer.clawbot.hk/docs - **API Reference:** https://memorylayer.clawbot.hk/api - **Community:** Discord (link in docs) - **Issues:** GitHub (link in docs) ## Links - **Homepage:** https://memorylayer.clawbot.hk - **Dashboard:** https://dashboard.memorylayer.clawbot.hk - **API Docs:** https://memorylayer.clawbot.hk/docs - **Python SDK:** https://pypi.org/project/memorylayer (when published)
將開發過程中的經驗、錯誤與修正記錄為 Markdown 檔案,供日後重用並提升為工作區記憶。
用自然語言描述需求,在本機與遠端技能市場尋找技能並取得推薦。
給 AI agent 用的瀏覽器自動化命令列技能。
將使用者明確糾正與自我檢視整理成分層記憶,並依使用情況升級或封存。
在安裝第三方技能之前,按清單審查來源、許可權與危險模式的技能。
將代理記憶與跨技能狀態整理成具型別與約束的知識圖譜。
規劃 AI 代理主動跟進、工作記憶與回顧流程的工作區設定指南。
指導如何建立與更新有效 Claude 技能的寫作指南。
資料月份:2026-09 · 下載/星標/安裝數聚合自公開技能註冊庫(ClawHub、SkillHub),按月更新。口徑說明見資料方法。