memro MCP Server
Provides persistent memory for AI models by enabling the storage and retrieval of episodic, semantic, and procedural information through the memro protocol. It allows assistants to maintain long-term context via semantic search and chronological memory management.
README
memro MCP Server
Gives any MCP-compatible AI (Claude Desktop, Cursor, Zed, etc.) persistent memory via the memro protocol.
Install (Local)
hhhhhhhhggg
pip install memro-mcp
Install (Docker)
ojkjdnksdnodk;lka;dljladnaljdnalkdnaldnadnkadnjkadnlkdjwk docker build -t memro-mcp .
Local Mode (Default)
docker run -e MEMRO_BASE_URL=http://your-backend-url memro-mcp
Cloud Mode (SSE)
docker run -p 8080:8080 -e MCP_TRANSPORT=sse -e MEMRO_BASE_URL=http://your-backend-url memro-mcp
## Managed Mode (SaaS)
For multi-tenant or SaaS deployments where you want a single MCP server instance to serve multiple users:
1. Start the server with `MCP_TRANSPORT=sse`.
2. The agent must call the `initialize_session` tool first to register its `agent_id` and `private_key`.
3. All subsequent calls in that session will use those credentials.
See [setup_managed.md](setup_managed.md) for a detailed configuration guide.
## Setup with Claude Desktop
Add to `~/.claude/claude_desktop_config.json`:
```json
{
"mcpServers": {
"memro": {
"command": "memro-mcp",
"env": {
"MEMRO_AGENT_ID": "your_agent_id_here",
"MEMRO_PRIVATE_KEY": "your_private_key_here",
"MEMRO_BASE_URL": "http://localhost:8081"
}
}
}
}
First run without
MEMRO_AGENT_IDset will auto-create a new agent and print the credentials to stderr.
Available Tools
Once connected, Claude (or any MCP client) can use:
| Tool | What it does |
|---|---|
remember |
Store a new memory (episodic, semantic, procedural, or profile) |
recall |
Semantic search across all memories |
get_recent_memories |
Get most recent memories chronologically |
delete_memory |
Delete a specific memory by ID |
export_memories |
Export a summary of all stored memories |
Example Usage
Once configured, Claude will automatically use memory:
You: Remember that I prefer TypeScript over JavaScript
Claude: [calls remember("User prefers TypeScript over JavaScript", type="profile")]
Got it, I'll remember that.
You: What's my preferred language?
Claude: [calls recall("preferred programming language")]
Based on what I remember, you prefer TypeScript over JavaScript.
Self-host the memro backend
cd memrohq
docker-compose up -d
# Backend runs at http://localhost:8081
Environment Variables
| Variable | Required | Default | Description |
|---|---|---|---|
MEMRO_AGENT_ID |
No* | auto-created | Your agent's public key |
MEMRO_PRIVATE_KEY |
No* | auto-created | Your agent's private key |
MEMRO_BASE_URL |
No | http://localhost:8081 |
memro backend URL |
*First run will auto-create and print credentials. Save them for future runs.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。