Skip to content

9. ZTM‐MCP‐Remote: Intro & Quick Start

Jump to bottom
CaiShu edited this page Jun 18, 2025 · 1 revision

Flomesh MCP Remote (FMR) bridges the gap between MCP Clients and Servers, dramatically simplifying AI Agent development and delivering powerful, enterprise-grade network infrastructure to effortlessly achieve robust zero-trust security.

Quick Links：

  • Developer Quick Start
  • Run the MCP-Remote proxy on server side and map Stdio to local
  • Apply Access Control
  • Build Multi-MCP and Multi-LLM mesh network and ready for zero trust

Developer Quick Start

  1. Download ztm app from GitHub nightly build page : https://github.com/flomesh-io/ztm/actions/runs/15553484715/artifacts/3294106837

  2. Double click the dmg file and install

  3. Start ztm app from application menu

  4. Click LLM/MCP config button and fill in parameters & configurations menu llm-config

Config Deepseek: deep-seek-config

Config Excel MCP Server: mcp-server-config

  1. Open ztm Chat and setup LLM and MCP to work with Access http://localhost:7777 in web browser: open-chat-in-browser

Click 'Message' on left bar , then click 'AI Bot': chatbot

Set which LLM and MCP to use for this chat session: setup-chat

Chat and ask to create an Excel file: prompt-1 prompt-2

And check the files created in excel-mcp-server directory (where we run the excel mcp server).

  1. MCP replay Click on the timer icon on right up corner, we can see how chat agent interact with LLM and how LLM call MCP tools and detail steps/parameters during the whole process: mcp-replay

  2. About local RESTful Back to LLM/MCP config, we can see there are two 'Routes': local-restful

Behinds the screen, Flomesh MCP Remote (FMR) aggregates all LLM & MCP services and map them to local URL, it's called 'local RESTful' route, and FMP creates these local RESTful route automatically. Developers can easily interact with these local RESTful with their favorite dev tools like vscode and curl.

Clone this wiki locally