About polydev
polydev is a multi-model consultation MCP server for coding agents. It queries GPT-5, Claude, Gemini, and Grok together and returns synthesized perspectives, with hosted and self-hosted options.
Key features
- 01Queries GPT-5, Claude, Gemini, and Grok together
- 02Synthesizes perspectives from multiple models
- 03Offers hosted and self-hosted options
- 04Can route requests through CLI subscriptions
Use cases
- 01Get varied opinions on a coding task that is stuck
- 02Assess refactoring or technology choices
MCP support
The MCP server exposes the get_perspectives tool, which accepts a prompt and returns combined model perspectives. It can be configured with Claude Code, Cursor, Windsurf, Cline, and OpenAI Codex CLI.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
npx -y polydev-aiPOLYDEV_USER_TOKENSecretYour Polydev API token from polydev.ai/dashboard/mcp-tokensLocal servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.
Tools
Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.
Details
From the project READMEPolydev
Multi-model AI perspectives for your coding agents
Query GPT-5, Claude, Gemini, and Grok simultaneously through one MCP server
🚀 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
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:
{
"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):
{
"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:
[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:
{
"tool": "get_perspectives",
"arguments": {
"prompt": "How should I refactor this authentication flow?"
}
}
Or just mention "polydev" or "perspectives" in your prompt:
"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 |
Project Structure
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.
Links
- Website: polydev.ai
- Documentation: polydev.ai/docs
- npm: npmjs.com/package/polydev-ai
- Research: SWE-bench Paper
License
MIT License - see LICENSE for details.
Built by Polydev AI
Multi-model consultation for better code
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.