MCP Servers

模型上下文协议服务器、框架、SDK 和模板的综合目录。

MCP server for Origin. Where understanding compounds.

创建于 4/19/2026
更新于 about 2 hours ago
Repository documentation and setup instructions

origin-mcp

MCP server for Origin. Where understanding compounds.

Connects AI tools (Claude Code, Cursor, Claude Desktop, ChatGPT, Gemini CLI, Windsurf) to Origin's memories, concepts, decisions, and knowledge graph via the Model Context Protocol.

Install

cargo install origin-mcp

Or via npm (downloads the binary automatically):

{
  "mcpServers": {
    "origin": {
      "command": "npx",
      "args": ["-y", "origin-mcp"]
    }
  }
}

Usage

  1. Start the Origin daemon (origin-server) or desktop app
  2. Configure your AI tool with the MCP config above

The server connects to Origin via HTTP (127.0.0.1:7878).

Options

--origin-url <URL>    Override Origin server URL

Tools

| Tool | Description | Annotations | |------|-------------|-------------| | remember | Store a memory, fact, preference, or decision | write, non-destructive | | recall | Search memories and knowledge graph | read-only | | context | Load session context (identity, preferences, goals, relevant memories) | read-only | | forget | Delete a memory and clean up entity links | destructive, idempotent |

Agent guidance

The server ships with proactive-capture instructions for agents, including a two-model mental framing (profile = about the user; knowledge = about the world) and anti-noise rules. Agents should generally omit memory_type and let the backend auto-classify. See src/tools.rs for inline tool schemas and with_instructions text.

License

MIT

快速设置
此服务器的安装指南

安装命令 (包未发布)

git clone https://github.com/7xuanlu/origin-mcp
手动安装: 请查看 README 获取详细的设置说明和所需的其他依赖项。

Cursor 配置 (mcp.json)

{ "mcpServers": { "7xuanlu-origin-mcp": { "command": "git", "args": [ "clone", "https://github.com/7xuanlu/origin-mcp" ] } } }