How do I use Augment-MCP-Proxy?

1. Click on "Deploy Server". 2. Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state. 3. In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Augment-MCP-Proxy Switch to claude-3-opus and analyze my current project structure." That's it! The server will respond to your query, and you can continue using it as needed. Here is a step-by-step guide with screenshots.

de en es ja ko ru zh

Augment-MCP-Proxy

by Windyhhh

Overview Schema Source Related Servers Discussions Admin

Local

🔌 Augment MCP Proxy Server | Augment MCP Proxy

Use Third-Party Anthropic API in Augment — Save Your Credits
A lightweight MCP proxy server that connects Augment (Anthropic's official VS Code extension) to third-party Anthropic APIs, so Augment credits are not consumed — only third-party API quota is used. It runs fully locally, with code analysis done on your machine, keeping your privacy safe and secure.
A lightweight MCP proxy server that connects Augment (Anthropic's official VS Code extension) to third-party Anthropic APIs, saving Augment credits. Runs fully locally with code analysis done on your machine for privacy.

✨ Key Features

Feature	Description
💰 Save Credits	No Augment credits consumed; only third-party API quota used
🔄 Seamless Integration	Augment's experience stays exactly the same; all features supported
🔒 Runs Locally	Code analysis happens locally, source code is never uploaded, privacy-safe
🤖 Multi-Model Support	Switch freely among three models: Haiku / Sonnet / Opus
⚡ Lightweight	Only 2 dependencies (axios + dotenv), starts in seconds
🪟 Cross-Platform	Runs on Windows / macOS / Linux

Related MCP server: roblox-mcp-bridge

🏗️ How It Works

┌──────────────┐     HTTP      ┌──────────────────┐     HTTPS      ┌──────────────────┐
│   Augment    │ ────────────→ │  MCP Proxy       │ ─────────────→ │  Third-Party     │
│   (VS Code)  │ ←──────────── │  (localhost:3000)│ ←───────────── │  Anthropic API   │
└──────────────┘    响应       └──────────────────┘    响应         └──────────────────┘
                                    │
                                    ├─ /v1/models    → 模型列表
                                    ├─ /v1/messages  → 消息转发
                                    └─ /health        → 健康检查

The Augment plugin connects to the local proxy server via the MCP protocol. The proxy forwards requests to the third-party Anthropic API and returns responses to Augment. The entire process is completely transparent to Augment.

🚀 Quick Start

Prerequisites

Node.js >= 14.0
npm >= 6.0

Installation Steps

# 1. 克隆项目
git clone https://github.com/Windyhhh/Augment-MCP-Proxy.git
cd Augment-MCP-Proxy

# 2. 安装依赖
npm install

# 3. 配置环境变量
cp .env.example .env
# 编辑 .env，填入你的第三方 API Key

Configure `.env`

# 第三方 API 地址（兼容 Anthropic 格式的 API）
THIRD_PARTY_API=https://your-api-endpoint.com/anthropic

# 你的 API Key
API_KEY=sk-your-api-key-here

# MCP 服务器端口（默认 3000）
MCP_PORT=3000

Start the Server

# 方式一：npm
npm start

# 方式二：直接运行
node index.js

# 方式三（Windows）：PowerShell 脚本
.\start-server.ps1 start

Once it starts successfully, you will see:

========================================
MCP 代理服务器启动成功！
========================================
服务器地址：http://localhost:3000
第三方 API：https://your-api-endpoint.com/anthropic
支持的模型：
  - claude-3-haiku-20240229 (Claude 3 Haiku)
  - claude-3-sonnet-20240229 (Claude 3 Sonnet)
  - claude-3-opus-20240229 (Claude 3 Opus)
========================================

⚙️ Configuring Augment

1. Locate the MCP config file

Windows:

%APPDATA%\Code\User\globalStorage\anthropic.augment\mcp-servers.json

macOS:

~/Library/Application Support/Code/User/globalStorage/anthropic.augment/mcp-servers.json

Linux:

~/.config/Code/User/globalStorage/anthropic.augment/mcp-servers.json

2. Write the Configuration

Copy the contents of mcp-servers.json to the file above:

{
  "mcpServers": {
    "augment-proxy": {
      "command": "node",
      "args": ["path/to/index.js"],
      "env": {
        "MCP_PORT": "3000"
      }
    }
  }
}

3. Restart VS Code

Completely close VS Code and reopen it, then wait 5–10 seconds for the MCP server to load.

4. Get Started

Enter the following in Augment's chat:

use claude-3-haiku-20240229, 你好！请给我一个 Python Hello World 示例

🧪 Testing & Verification

Health Check

curl http://localhost:3000/health

Expected output:

{"status":"ok","timestamp":"2025-01-01T00:00:00.000Z"}

Fetch the Model List

curl http://localhost:3000/v1/models

Run the Test Script (Windows)

.\test-api.ps1

Confirm the Credit Savings

Compare Augment's credit balance before and after using the MCP proxy; it should not decrease.

📁 Project Structure

Augment-MCP-Proxy/
├── README.md                  # 本文件
├── package.json               # 项目配置
├── package-lock.json          # 依赖锁定
├── .env.example               # 环境变量模板
├── .gitignore                 # Git 忽略规则
├── index.js                   # MCP 代理服务器主文件
├── start-server.ps1           # Windows 启动/停止/重启脚本
├── test-api.ps1               # API 测试脚本
├── mcp-servers.json           # Augment MCP 配置示例
└── AUGMENT_CONFIG_GUIDE.md    # Augment 详细配置指南

🔧 API Endpoints

`GET /health`

Health check endpoint.

Response:

{
  "status": "ok",
  "timestamp": "2025-01-01T00:00:00.000Z"
}

`GET /v1/models`

Retrieves the list of supported models.

Response:

{
  "object": "list",
  "data": [
    {
      "id": "claude-3-haiku-20240229",
      "object": "model",
      "created": 1735689600,
      "owned_by": "anthropic"
    }
  ]
}

`POST /v1/messages`

Forwards message requests to third-party APIs.

Request body:

{
  "model": "claude-3-sonnet-20240229",
  "max_tokens": 4096,
  "messages": [
    {"role": "user", "content": "Hello!"}
  ],
  "temperature": 0.7
}

🎯 Supported Models

Model	Speed	Quality	Cost	Best Use Cases
Haiku	⚡⚡⚡	⭐⭐	Lowest	Simple questions, code completion, everyday chat
Sonnet	⚡⚡	⭐⭐⭐⭐	Medium	General tasks, code review, recommended default
Opus	⚡	⭐⭐⭐⭐⭐	Higher	Complex problems, architecture design, deep analysis

To switch models in Augment:

use claude-3-haiku-20240229, ...
use claude-3-sonnet-20240229, ...
use claude-3-opus-20240229, ...

🛠️ PowerShell Script Usage

# 启动服务器
.\start-server.ps1 start

# 停止服务器
.\start-server.ps1 stop

# 重启服务器
.\start-server.ps1 restart

# 测试连接
.\start-server.ps1 test

This server cannot be deployed

RayrunBuild, deploy, and operate MCP servers without becoming an infrastructure team.ray.runAd

Maintenance

ActivityMaintained

ResponsivenessNo issues

Resources

GitHub Repository

Need Help?

Related Servers

Related MCP Connectors

SprawlOAuth
to.sprawl
Build and supervise fleets of agents from Claude Code, Codex or Cursor. Connects over OAuth.
Agent Orchestration Autonomous Agents App Automation
1
QuallaaOAuth
com.quallaa
Talk to your public-facing AI from any MCP client — Claude, ChatGPT, Cursor, Cline, Windsurf.
AI & Machine Learning Communication
2
OpenRouter AOAuth
ai.openrouter.mcp
The OpenRouter MCP server plugs OpenRouter into the AI tools you already use. Once connected, your assistant can pull live OpenRouter data (models, prices, your credits, rankings, and docs) and send quick test messages, all without leaving your editor.
AI & Machine Learning
221
branchly A
io.branchly
Connect your agent to branchly: manage content, prompts, actions, learn from analytics and insights
Agent Orchestration AI & Machine Learning Content Management Systems
371

Related MCP Servers

LiteLLM MCP Server Bridge
AI & Machine Learning Text-to-Speech
AbhishekMore-1
A
license
B
quality
C
maintenance
Connects Antigravity to a LiteLLM instance, enabling chat completions, model listing, health checks, and more via the Model Context Protocol.
Updated 6 months ago (2026-03-02 22:20 UTC)
11
23
6
MIT
roblox-mcp-bridge
kavinr24
A
license
Not graded
quality
B
maintenance
Connects Roblox Studio to local AI agents via HTTP, allowing them to control Studio tools.
Updated 2 months ago (2026-06-29 17:21 UTC)
MIT
Coolify AnythingLLM MCP
Cloud Platforms CI/CD & DevOps
xlop-dev
A
license
Not graded
quality
C
maintenance
Connects AnythingLLM to Coolify for natural language management of self-hosted infrastructure, including diagnostics, log fetching, and deployment automation.
Updated 2 months ago (2026-07-24 11:44 UTC)
MIT
everos-mcp
Knowledge & Memory Search
Marvisatron
A
license
A
quality
C
maintenance
Connects Claude Code to a local EverOS instance for searching and storing long-term memory, keeping data entirely on your machine.
Updated 2 months ago (2026-07-08 04:26 UTC)
4
Apache 2.0