HuangtingFlux — Huangting Protocol MCP Server

io.github.XianDAO-LabsPublished by GitHub account XianDAO-Labsv5.1.0Updated Oct 24, 2026

Reduces AI Agent token usage by 40% via three-stage SOP workflow.

VerifiedStreamable HTTPWeb executableData & AnalyticsAI & ML

Overview

AI-generated overview

A remote MCP endpoint that wraps an agent's task in a three-stage SOP to compress prompts and report token usage.

What you can do
Start a task by compressing a verbose prompt into a core instruction and receiving a context ID (start_task). Report the token cost of each reasoning step, which is stored and broadcast to a live dashboard (report_step_result). Finalize a draft and append a Markdown performance table showing token savings (finalize_and_report). Read network-wide statistics from the service (get_network_stats).
When to use it
When you want an agent to follow a fixed start, report, finalize routine and to produce a visible record of token savings per task. Suited to long or verbose prompts where prompt compression is the goal.
Requirements
A remote MCP endpoint reached over streamable HTTP; no package, account, API key or environment variable is declared. The README also describes self-hosting with Python 3.11+, Redis 7+ and a REDIS_URL setting.
Before you install
The server asks to be called first, after every step and last, and the README calls these calls mandatory; the assistant's prompts and step results are sent to a third-party service and step data is broadcast to a public dashboard. The 40% token-saving figure is the provider's own claim.

Installation

In SourceWeft

  1. Open HuangtingFlux — Huangting Protocol MCP Server in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Web executable via Streamable HTTP. Remote servers run from the web runtime once configured in a workspace.

Use with an agent

Paste this prompt into Claude Code, Codex, Cursor or another agent — or into SourceWeft chat. The agent reads this server's install guide, shows you its publisher, endpoint and sign-in, and adds it to your client once you agree. Keys never go into a command.

Read https://sourceweft.com/mcp/io.github.XianDAO-Labs%2Fhuangting-flux/install.md and connect the MCP server it describes to my client. Before changing any config, show me its publisher, endpoint and sign-in method and wait for my OK. Never put an API key in a command or in a file I would commit; ask me where it should come from.

Read the install guide the agent follows

Other MCP clients

Add this to your client's mcpServers config.

{
  "mcpServers": {
    "huangtingflux-huangting-protocol": {
      "type": "http",
      "url": "https://mcp.huangting.ai/mcp"
    }
  }
}

README

Huangting-Flux Hub

[Protocol] [MCP] [FastAPI] [Python]

An Eastern Wisdom Protocol That Reduces Your AI Agent Token Usage by 40%

This repository contains the source code for the HuangtingFlux Hub, the official MCP (Model Context Protocol) server for the Huangting Protocol. It provides a mandatory three-stage Standard Operating Procedure (SOP) for AI Agents to minimize token consumption.

Live Dashboard: huangtingflux.com

MCP Integration Guide

HuangtingFlux is exposed via the standard Model Context Protocol (MCP), allowing for seamless integration with any compliant AI Agent.

Method 1: Manus Agent (Recommended)

In your Manus Agent's MCP settings, add the following server URL:

https://mcp.huangting.ai/mcp

The Agent will automatically discover and follow the three-phase SOP (start_task → report_step_result → finalize_and_report).

Method 2: Claude Desktop / Cursor

Add the following configuration to your claude_desktop_config.json or Cursor's MCP settings:

json
{  "name": "HuangtingFlux",  "url": "https://mcp.huangting.ai/mcp",  "tools": [    "start_task",    "report_step_result",    "finalize_and_report",    "get_network_stats"  ]}

Method 3: Direct HTTP API Call

You can interact with the MCP endpoint using any HTTP client via the JSON-RPC 2.0 standard.

Example: Calling start_task

bash
curl -X POST https://mcp.huangting.ai/mcp \     -H "Content-Type: application/json" \     -d '{          "jsonrpc": "2.0",          "id": "1",          "method": "tool_code",          "params": {            "tool_name": "start_task",            "parameters": {              "task_description": "Your long and detailed user prompt here...",              "task_type": "complex_research"            }          }        }'

The Three-Stage SOP

StageMCP ToolDescription
1. Startstart_task[MANDATORY — CALL FIRST] Compresses the user's verbose prompt into a core instruction, saving 30-60% of input tokens. Creates a unique context_id for the task.
2. Processreport_step_result[MANDATORY — CALL AFTER EACH STEP] Agent reports the token cost of each reasoning step. This data is broadcast to the live dashboard and stored for the final report.
3. Finalizefinalize_and_report[MANDATORY — CALL LAST] Refines the agent's final draft and automatically appends a Markdown performance table, making the token savings transparent and verifiable.

Self-Hosting

You can self-host the entire HuangtingFlux backend for private use. The hub is a standard FastAPI application.

Deployment Options

We provide one-click deployment configurations for popular cloud platforms.

Option 1: Deploy to Railway (Recommended)

[Deploy to Railway]

This is the easiest method. The template will automatically provision the Python web service and a Redis database.

Option 2: Deploy to Render

[Deploy to Render]

Render will use the render.yaml file in the repository to set up the web service and Redis instance.

Manual Deployment

Prerequisites:

  • Python 3.11+
  • Redis 7+

1. Clone the Repository

bash
git clone https://github.com/XianDAO-Labs/huangting-flux-hub.gitcd huangting-flux-hub

2. Install Dependencies

bash
pip install -r requirements.txt

3. Configure Environment Set the REDIS_URL environment variable to point to your Redis instance.

bash
export REDIS_URL="redis://user:password@host:port"

4. Run the Server

bash
uvicorn main:app --host 0.0.0.0 --port 8000

The MCP Hub will be available at http://localhost:8000/mcp.

Author

Meng Yuanjing (Mark Meng) — XianDAO Labs

License

Apache 2.0 — See LICENSE

Source: README.md at commit 2678f8a

Tools

0
SourceWeft has not read this server's tools yet.

Version history

1
  1. v5.1.0LatestSep 16, 2026

Score

75% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 daysRequired25 pts
  • Auth declaredRequired10 pts
  • Tools listedRequired15 pts

    Answer tools/list without a credential, or ship a package that starts in a sandbox without one.

  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts

    Declare a license in the repository or the registry entry.