ai.memcell由 memcell.ai 發布官方發布者v2.2.1更新於 Nov 7, 2026
Shared, persistent memory for AI agents: recall before acting, report outcomes to earn confidence.
MemCell 為 AI 助理提供共享的持久工作區記憶：行動前召回指令與事實，事後回報結果以校準信心。
Web executable，透過 Streamable HTTP。 遠端服務在工作區中設定後即可從網頁執行環境執行。
把這段提示詞貼上到 Claude Code、Codex、Cursor 等代理程式，或 SourceWeft 對話中。代理程式會讀取該伺服器的安裝指南，先向你展示發布者、端點和登入方式，經你同意後再把它新增到你的客戶端。金鑰不會寫進任何命令。
閱讀 https://sourceweft.com/mcp/ai.memcell%2Fmemcell/install.md，並按其中說明把這個 MCP 伺服器接入我的客戶端。修改任何設定之前，先告訴我它的發布者、端點和登入方式，等我確認。不要把 API 金鑰寫進命令或我會提交的檔案；問我金鑰應從哪裡取得。把它新增到你客戶端的 mcpServers 設定中。
{
"mcpServers": {
"memcell": {
"type": "http",
"url": "https://memcell.ai/mcp",
"headers": {
"Authorization": "Bearer <YOUR_TOKEN>"
}
}
}
}
The official Model Context Protocol (MCP) server & registry manifest for MemCell.
Living, persistent memory for AI agents: recall before acting, report outcomes so confidence is earned.
[License: Apache-2.0] [Documentation] [Smithery]
MemCell provides persistent, adaptive memory, working thread continuity, knowledge graph traversal, and grounded document RAG for AI agents across any client harness. This repository contains the official MCP registry manifest, Smithery configuration, and connection guides.
Run via npx (requires memcell CLI installed or accessible via npm):
Connect directly to the hosted or self-hosted MemCell streamable HTTP endpoint:
https://memcell.ai/mcpAuthorization: Bearer <agent-key>Run MemCell on your local machine with embedded PGlite vector storage (once cached or pointed at MEMCELL_SERVER_PATH, memcell start runs completely offline without network access):
Add to claude_desktop_config.json:
Run:
Add to .cursor/mcp.json (or ~/.cursor/mcp.json):
Add to ~/.gemini/antigravity/mcp_config.json:
Configure the MCP server command as:
npx["-y", "memcell", "mcp"]MemCell exposes seven workspace tools over MCP:
recall: Pre-flight memory retrieval before taking action. Fetches relevant instructions, decisions, facts, preferences, active or resumed session threads, 1-hop knowledge graph relations, and grounded workspace document chunks.remember: Durable memory storage. Files atomic instructions, decisions, facts, and preferences—including detailed operational specifications (content), affected resources (resources), and enforcement moments (appliesAt)—directly into the active workspace or user profile.report: Outcome feedback calibration. Reports whether recalled memory worked, failed, or was avoided, calibrating confidence dynamically over time.feedback: Direct evaluation of a specific memory ID (worked or failed) with supporting outcome notes.ingest_document: Multi-format document ingestion. Indexes PDF, DOCX, CSV, TSV, HTML, JSON, and Markdown documents into grounded workspace chunks and optionally distills atomic memories.search_documents: Hybrid semantic and full-text retrieval across grounded workspace document chunks.get_insights: Live workspace telemetry, empirical Confidence Calibration Curve (ECE, Brier score, and bucketed observed success rates), and Governance Shield metrics (enforced guards, quarantined user-scope memories, and demoted items).Building custom agents in TypeScript or Python? Official drop-in adapters ship inside @memcell/sdk (@memcell/sdk/vercel, @memcell/sdk/openai, @memcell/sdk/langgraph) and Python memcell.integrations (OpenAI Agents, LangGraph, PydanticAI, CrewAI) alongside MCP.
Apache-2.0
來源：README.md，提交 c273c96
Ask the workspace's memory before acting. Evaluates intent against durable instructions, decisions, facts, preferences, active session threads, and grounded RAG documents. Zero conversation retention: query text is strictly ephemeral and never stored. Pass the recallId to 'report' so task outcomes sharpen what was recalled.
tagstypelimitqueryscopeintentscopesmetadataRecord knowledge into MemCell. Accepts raw text, conversational excerpts, or structured hints. MemCell distills material into atomic memories, links them into the workspace Knowledge Graph, and checkpoints session threads.
rawtypescopetitlecontentcontextlearningClose the recall loop. Reports whether a recalled memory worked, failed, or was avoided in practice. Strengthens what worked and surfaces what failed for correction.
reasonoutcome必填recallIdmemory_idrecall_idaction_takenEvaluate a specific memory after execution: report whether its advice worked or failed in practice. Dynamically updates confidence and appends evidence.
reasonoutcome必填memory_id必填recall_idIngest a PDF, DOCX, CSV, HTML, JSON, or Markdown document into the workspace's grounded knowledge base and optionally distill atomic memories.
titlecontentdistillfilenamemimeTypefileBase64Search grounded workspace document chunks (PDF, DOCX, CSV, HTML, JSON, Markdown) using hybrid semantic and full-text retrieval.
limitquery必填Inspect workspace telemetry, empirical confidence calibration curve (predicted confidence vs observed worked rate, ECE, and Brier score), and governance shield metrics (enforced guards, quarantined user-scope memories, and demoted items).
timeframe已滿足適用檢查項的 100%
該評分衡量此收錄條目的完整程度，依據註冊中心條目、端點與公開來源計算得出。它不是對伺服器本身的評價，也不表示伺服器安全或目前可用：可連通性是帶日期的檢查結果，評分也不包含安裝量或使用者評價。
npx -y memcell mcp# 1. Start local server in backgroundmemcell start --daemon
# 2. Wire local MCP toolsmemcell connect --url http://localhost:3000{ "mcpServers": { "memcell": { "command": "npx", "args": ["-y", "memcell", "mcp"] } }}claude mcp add memcell -- npx -y memcell mcp{ "mcpServers": { "memcell": { "command": "npx", "args": ["-y", "memcell", "mcp"] } }}{ "mcpServers": { "memcell": { "command": "npx", "args": ["-y", "memcell", "mcp"] } }}