io.github.polydev-ai/polydev

Name: io.github.polydev-ai/polydev
Author: polydev-ai

AI 与智能体

by polydev-ai

通过一个 MCP server 同时查询 GPT-5、Claude、Gemini 和 Grok，统一访问多模型能力。

GitHub

什么是 io.github.polydev-ai/polydev？

通过一个 MCP server 同时查询 GPT-5、Claude、Gemini 和 Grok，统一访问多模型能力。

README

Polydev

<img src="https://raw.githubusercontent.com/polydev-ai/polydev/main/public/logo.png" alt="Polydev Logo" width="120" /> Multi-model AI perspectives for your coding agents Query GPT-5, Claude, Gemini, and Grok simultaneously through one MCP server <a href="https://www.npmjs.com/package/polydev-ai"><img src="https://img.shields.io/npm/v/polydev-ai.svg" alt="npm version"></a> <a href="https://polydev.ai/articles/swe-bench-paper"><img src="https://img.shields.io/badge/SWE--bench-74.6%25-brightgreen" alt="SWE-bench Verified"></a> <a href="https://opensource.org/licenses/MIT"><img src="https://img.shields.io/badge/License-MIT-blue.svg" alt="License: MIT"></a>

🚀 One-Command Install

Claude Code

claude mcp add polydev -- npx -y polydev-ai@latest

Cursor / Windsurf / Cline

npx polydev-ai@latest

OpenAI Codex CLI

npx polydev-ai@latest

Get your token at polydev.ai/dashboard/mcp-tokens

Why Polydev?

Stop copy-pasting between ChatGPT, Claude, and Gemini. Get all their perspectives in your IDE with one request.

Metric	Result
SWE-bench Verified	74.6% Resolve@2
Cost vs Claude Opus	62% lower
Response time	10-40 seconds

"Different models have different blind spots. Combining their perspectives eliminates yours."

How It Works

code

Your Agent → Polydev MCP → [GPT-5, Claude, Gemini, Grok] → Synthesized Answer

When your AI agent gets stuck, Polydev consults multiple frontier models simultaneously and returns their combined perspectives. One request, four expert opinions.

Quick Start

Option 1: Use Hosted Service (Recommended)

Get started instantly at polydev.ai

Step 1: Install the MCP server

npx polydev-ai@latest

Step 2: Set your token (get it from polydev.ai/dashboard/mcp-tokens)

export POLYDEV_USER_TOKEN="pd_your_token_here"

Option 2: Self-Host with Your Own API Keys

Step 1: Clone and install

git clone https://github.com/polydev-ai/polydev.git

cd polydev && npm install

Step 2: Configure environment

cp .env.example .env.local

Step 3: Add your API keys to .env.local and run

npm run dev

IDE Configuration

Claude Code

One command:

claude mcp add polydev -- npx -y polydev-ai@latest

Or add to ~/.claude.json:

json

{
  "mcpServers": {
    "polydev": {
      "command": "npx",
      "args": ["-y", "polydev-ai@latest"],
      "env": {
        "POLYDEV_USER_TOKEN": "pd_your_token_here"
      }
    }
  }
}

Cursor / Windsurf / Cline

Add to your MCP config (usually ~/.cursor/mcp.json or similar):

json

{
  "mcpServers": {
    "polydev": {
      "command": "npx",
      "args": ["-y", "polydev-ai@latest"],
      "env": {
        "POLYDEV_USER_TOKEN": "pd_your_token_here"
      }
    }
  }
}

OpenAI Codex CLI

Add to ~/.codex/config.toml:

toml

[mcp_servers.polydev]
command = "npx"
args = ["-y", "polydev-ai@latest"]

[mcp_servers.polydev.env]
POLYDEV_USER_TOKEN = "pd_your_token_here"

[mcp_servers.polydev.timeouts]
tool_timeout = 180
session_timeout = 600

Usage

Once connected, your agent can call:

json

{
  "tool": "get_perspectives",
  "arguments": {
    "prompt": "How should I refactor this authentication flow?"
  }
}

Or just mention "polydev" or "perspectives" in your prompt:

code

"Use polydev to debug this infinite loop"
"Get perspectives on: Should I use Redis or PostgreSQL for caching?"

Use Your Existing CLI Subscriptions

Already paying for ChatGPT Plus, Claude Pro, or Gemini Advanced? Use those subscriptions directly through your CLI tools.

Subscription	CLI Tool	Setup
Claude Pro ($20/mo)	Claude Code	`claude login`
ChatGPT Plus ($20/mo)	Codex CLI	`codex login`
Gemini Advanced ($20/mo)	Gemini CLI	`gemini login`

Polydev can route requests through your authenticated CLI sessions — your subscription quota is used, no extra API costs.

Self-Hosting

Requirements

Node.js 18+
PostgreSQL (or Supabase)
API keys for the models you want to use

Environment Variables

Create a .env.local file:

# Database
NEXT_PUBLIC_SUPABASE_URL=your_supabase_url
NEXT_PUBLIC_SUPABASE_ANON_KEY=your_anon_key
SUPABASE_SERVICE_ROLE_KEY=your_service_role_key

# AI Providers (add the ones you want)
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
GOOGLE_API_KEY=...
XAI_API_KEY=...

Database Setup

cd supabase && supabase db push

Run

npm run dev

Research

Our approach achieves 74.6% on SWE-bench Verified (Resolve@2), matching Claude Opus at 62% lower cost.

Approach	Resolution Rate	Cost/Instance
Claude Haiku (baseline)	64.6%	$0.18
+ Polydev consultation	66.6%	$0.24
Resolve@2 (best of both)	74.6%	$0.37
Claude Opus (reference)	74.4%	$0.97

Read the full paper →

Project Structure

code

polydev/
├── src/                    # Next.js application
│   ├── app/               # App router pages & API routes
│   ├── components/        # React components
│   ├── hooks/             # Custom React hooks
│   └── lib/               # Utilities and services
├── mcp/                   # MCP server implementation
├── supabase/              # Database migrations
├── docs/                  # Documentation
└── public/                # Static assets

Contributing

Contributions are welcome! Please read our contributing guidelines before submitting PRs.

# Fork the repository, then:
git clone https://github.com/YOUR_USERNAME/polydev.git

cd polydev && npm install

git checkout -b feature/amazing-feature

git commit -m 'Add amazing feature'

git push origin feature/amazing-feature

Then open a Pull Request.

License

MIT License - see LICENSE for details.

Built by <a href="https://polydev.ai">Polydev AI</a> Multi-model consultation for better code

常见问题