waggle-sensor/sage-mcp

本页由 AI 翻译，技术细节请以原文为准。查看原文

2 stars 5 forks 质量 66 趋势 12

概览

A Model Context Protocol server for the Sage Grande Testbed — cyberinfrastructure for AI@Edge. It exposes ~30 MCP tools, resources, and prompts for querying sensor data, submitting edge-computing jobs, discovering plugins, and browsing Sage documentation from any MCP-compatible client (Claude Desktop, Cursor, custom agents, etc.). Built on FastMCP 2.10+ and the MCP SDK 1.12+; supports stdio, sse, and streamable-http transports. There are three ways to run this — pick the one that fits. All three expose the exact same tools. The only difference between hosted and local is how much data you can reach: without Sage credentials, every tool still runs but falls back to . See Authentication & credentials for how to add your token. The server is already running at https://mcp.sagecontinuum.org/mcp. Point any MCP client at it — no install needed. (~/.cursor/mcp.json) or (~/Library/Application Support/Claude/claude_desktop_config.json on macOS): Get YOUR_ACCESS_TOKEN from https://portal.

README

以下为原文 README 节选，完整内容请查看原文。

Sage MCP Server

Built on FastMCP 2.10+ and the MCP SDK 1.12+; supports stdio, sse, and streamable-http transports.

Choose your setup

There are three ways to run this — pick the one that fits.

#	Path	Who it’s for	You need
1	Hosted (mcp.sagecontinuum.org)	Everyone. Zero install.	An MCP client + a Sage token.
2	Local — stdio	IDE users who want to run the server themselves.	Python 3.11+ and this repo.
3	Local — HTTP	Devs, custom agents, curl testing.	Python 3.11+ or Docker.

All three expose the exact same tools. The only difference between hosted and local is how much data you can reach: without Sage credentials, every tool still runs but falls back to public data only. See Authentication & credentials for how to add your token.

1. Hosted (mcp.sagecontinuum.org)

The server is already running at https://mcp.sagecontinuum.org/mcp. Point any MCP client at it — no install needed.

Cursor (~/.cursor/mcp.json) or Claude Desktop (~/Library/Application Support/Claude/claude_desktop_config.json on macOS):

{
  "mcpServers": {
    "sage": {
      "url": "https://mcp.sagecontinuum.org/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_USERNAME:YOUR_ACCESS_TOKEN"
      }
    }
  }
}

Get YOUR_ACCESS_TOKEN from . The Bearer value must be in username:token form for anything beyond public data.

Restart your client, then ask natural questions:

“Show me temperature readings from node W023 in the last hour.” “Find nodes in Chicago with recent camera images.” “What’s the highest temperature recorded today across all nodes?”

2. Local — stdio (for IDE clients)

Run the server as a subprocess of your IDE. Nothing binds to a network port. This is the most private mode — data and credentials never leave your machine.

Install:

git clone https://github.com/waggle-sensor/sage-mcp.git
cd sage-mcp
pip install -r requirements.txt
cp .env.example .env       # edit and add SAGE_USER / SAGE_PASS if you have them

Wire it into Cursor/Claude Desktop:

{
  "mcpServers": {
    "sage-local": {
      "command": "/absolute/path/to/sage-mcp/scripts/run-local-stdio.sh",
      "env": {

在 GitHub 查看完整 README

安装

This server does not publish a one-line install command.

Open the repository installation guide 新手？先看 mcpskill.org 教程 →

配置

{
  "mcpServers": {
    "sage": {
      "url": "https://mcp.sagecontinuum.org/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_USERNAME:YOUR_ACCESS_TOKEN"
      }
    }
  }
}