Engram

app.getengram由 getengram.app 發布官方發布者v0.1.1更新於 Oct 24, 2026

Persistent, verbatim, searchable memory for AI assistants — one memory across every MCP client.

已驗證Streamable HTTP可網頁執行知識與記憶AI 與機器學習

概覽

AI 產生的概覽

Engram 為 AI 助理提供持久、逐字保存的對話記憶,可跨工作階段與跨 MCP 用戶端進行語意搜尋。

可以做什麼
Engram 是遠端 MCP 記憶伺服器,完整、不壓縮地保存對話記錄,並透過嵌入向量依語意檢索。它提供六個工具:create_conversation、append_messages、search、get_conversation、list_conversations 與 delete_conversation。記憶會在已連線的 MCP 用戶端之間共用,助理可回想先前的工作階段。Claude Code 外掛另附一項技能,用來判斷何時檢索或儲存記憶。
適用情境
當你希望助理跨工作階段、跨不同 MCP 用戶端記住過去的對話,並依語意而非精確關鍵字找回相關脈絡時適用。適合能接受將完整對話記錄存放在託管服務的使用者。
執行需求
透過 streamable HTTP 連線至 mcp.getengram.app 遠端端點,無需本機套件或 API 金鑰。需要 OAuth 登入,登入過程中會建立免費帳號。僅支援 stdio 的用戶端需透過 npx 使用 mcp-remote 等橋接工具,這需要 Node.js。需要網路存取。
安裝前請注意
完整對話會逐字、不壓縮地存放在第三方託管服務上,聊天中的敏感內容會被保留。delete_conversation 工具會刪除對話及其資料。用量依方案計量:免費版每月上限 1,000 則訊息,另有付費的 Pro、Team 與 Enterprise 方案。

安裝

在 SourceWeft 中

  1. 開啟 儀表板中的 Engram,將其新增到工作區。
  2. 連接 所需的 憑證。它們會被私密儲存,絕不會在公開頁面上展示。
  3. 為需要使用其工具的對話啟用該服務。

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。

交給代理程式安裝

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/app.getengram%2Fengram/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。

檢視代理程式所依據的安裝指南

其他 MCP 客戶端

把它新增到你客戶端的 mcpServers 設定中。 客戶端會在首次連接時完成 OAuth 登入。

{
  "mcpServers": {
    "engram": {
      "type": "http",
      "url": "https://mcp.getengram.app/mcp"
    }
  }
}

README

Engram

Memory infrastructure for AI agents. Store every conversation verbatim. Search by meaning.

[Website] [Docs] [License]

Engram is an MCP-native memory server that stores complete, uncompressed conversation transcripts and makes them searchable via semantic search. Connect any MCP-compatible client — Claude Desktop, Claude Code, Cursor, Windsurf, Zed — and your agent remembers everything across sessions.

Quick start

Claude Code

/plugin marketplace add get-engram/engram/plugin install engram@engram

Restart Claude Code and approve access in the browser. There is no API key to copy and no config file to edit — the server implements the MCP authorization flow (RFC 9728 / 8414 / 7591, PKCE), so the client discovers it, registers itself, and signs you in. A free account is created as part of signing in.

The plugin also ships a skill that tells Claude when to search memory and what is worth saving, so context accumulates without being asked.

Any other MCP client

Point it at the remote server and let OAuth handle access:

json
{  "mcpServers": {    "engram": {      "type": "http",      "url": "https://mcp.getengram.app/mcp"    }  }}

For a client that only speaks stdio, bridge to the remote server with mcp-remote — it drives the same browser OAuth flow, so there is still no key to copy:

json
{  "mcpServers": {    "engram": {      "command": "npx",      "args": ["-y", "mcp-remote", "https://mcp.getengram.app/mcp"]    }  }}

(An older version of this README suggested npx @getengram/cli mcp; the CLI has no such command and that configuration never worked.)

How it works

  • Verbatim storage — every message stored exactly as sent, no summarization or compression
  • Semantic search — find relevant context by meaning using bge-base-en-v1.5 embeddings
  • MCP-native — speaks the Model Context Protocol natively, works with any compatible client
  • Multi-tenant — per-organization isolation, team seats, and API key management

Architecture

Runs entirely on Cloudflare's developer platform:

  • Workers — Hono.js API and MCP server
  • D1 — SQLite at the edge for messages and metadata
  • Vectorize — semantic search index
  • Workers AI — embedding generation

Read the full architecture deep-dive.

MCP tools

Engram exposes 6 tools via MCP:

ToolDescription
create_conversationStart a new conversation with optional title, tags, metadata
append_messagesAdd messages to an existing conversation
searchSemantic search across all conversations
get_conversationRetrieve a conversation with its messages
list_conversationsList conversations with filtering and pagination
delete_conversationRemove a conversation and its data

See the API reference for full parameters and examples.

Packages

PackageDescription
apps/mcp-serverCloudflare Worker — MCP server and REST API
apps/cliCLI and MCP bridge (@getengram/cli)
packages/sdkTypeScript SDK (@getengram/sdk)
packages/dbDatabase queries and migrations
packages/sharedShared constants, types, and utilities

Integration guides

Pricing

Memory is permanent. The limits below are how many messages a plan can hold, not a monthly quota: nothing expires, and deleting conversations frees space.

PlanPriceMemory
Free$010,000 messages
Pro$9/mo1,000,000 messages
Team$9/seat/mo, minimum 2 seats1,000,000 messages per seat, pooled
Education$3/mo per student, $3/seat for academic teamsPro or Team features with a verified university email
EnterpriseCustomUnlimited

View pricing

Links

License

Business Source License 1.1 — see LICENSE for details.

來源:README.md,提交 accbe3d

工具

0
該伺服器需要登入,因此 SourceWeft 無法在沒有帳號的情況下讀取其工具,也未收錄。已連接帳號所能看到的內容可能因帳號及其權限而異。

版本歷史

2
  1. v0.1.1最新Oct 13, 2026
  2. v0.1.0Sep 16, 2026

評分

已滿足適用檢查項的 75%

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。

  • 近 30 天內可連通必需25 分
  • 已宣告認證方式必需10 分
  • 已列出工具15 分

    請讓伺服器在無憑證時回應 tools/list,或發布可在沙箱中無憑證啟動的軟體包。

  • 描述必需15 分
  • 圖示10 分
  • 文件15 分
  • 授權條款10 分

    請在儲存庫或註冊中心條目中宣告授權條款。