Content
<p align="center">
<h1 align="center">AI Team MCP</h1>
<p align="center"><b>One prompt. Four AI minds. Better code.</b></p>
<p align="center">Claude writes code. ChatGPT designs architecture. Gemini reviews. Perplexity researches.<br>All inside one MCP server. All from your terminal.</p>
</p>
<p align="center">
<a href="https://github.com/needsomehelp/ai-team-mcp/stargazers"><img src="https://img.shields.io/github/stars/needsomehelp/ai-team-mcp?style=social" alt="Stars"></a>
<a href="https://github.com/needsomehelp/ai-team-mcp/actions"><img src="https://github.com/needsomehelp/ai-team-mcp/actions/workflows/ci.yml/badge.svg" alt="CI"></a>
<a href="#"><img src="https://img.shields.io/badge/Python-3.10+-blue.svg" alt="Python"></a>
<a href="#"><img src="https://img.shields.io/badge/Protocol-MCP-orange.svg" alt="MCP"></a>
<a href="LICENSE"><img src="https://img.shields.io/badge/License-MIT-green.svg" alt="License"></a>
<a href="#"><img src="https://img.shields.io/badge/API%20Keys-Not%20Required-red.svg" alt="No API Keys"></a>
</p>
<p align="center"><i>Stop coding with one AI. Start shipping with an entire AI engineering team.</i></p>
---
## Why Four AI Models?
Each AI has a different strength. AI Team MCP uses them all:
| AI | Role | What It Does Best |
|----|------|-------------------|
| **Claude** | Lead Coder | Implementation, debugging, refactoring |
| **ChatGPT** | Architect | System design, planning, trade-offs |
| **Gemini** | Reviewer | Code review, bugs, security analysis |
| **Perplexity** | Researcher | Latest docs, libraries, real-time info |
**All use your existing browser subscriptions. No API keys required.**
---
## How It Works
```
You: "Build a FastAPI auth system with JWT"
Perplexity (Researcher) "Latest FastAPI-Security + python-jose patterns..."
|
v
ChatGPT (Architect) "Here's the folder structure, middleware design..."
|
v
Claude (Lead Coder) *writes production-ready code*
|
v
Gemini (Reviewer) "Found 2 security issues, 1 optimization..."
|
v
Final Result Production code, reviewed and researched.
```
**One prompt. Four AI experts. Zero tab-switching.**
---
## Quick Start (2 minutes)
### Step 1: Clone & Install
```bash
git clone https://github.com/needsomehelp/ai-team-mcp.git
cd ai-team-mcp/ai-team
pip install -r requirements.txt
```
### Step 2: Add to Claude Code
```bash
claude mcp add ai-team -s user python3 $(pwd)/mcp_server.py
```
### Step 3: Log in your agents
```
> Use ai_team_login with service='chatgpt', token='<your access token>'
> Use ai_team_login with service='gemini', token='<your __Secure-1PSID cookie>'
> Use ai_team_login with service='perplexity', token='<your cookie string>'
```
### Step 4: Verify
```bash
claude mcp list
# Should show: ai-team ✓ Connected
```
Now just talk to Claude Code:
```
> "Ask ChatGPT to design a caching layer"
> "Have Gemini review this code"
> "Research the latest Stripe API with Perplexity"
> "Run the full AI team on building auth"
```
---
## Without AI Team vs With AI Team
| Without AI Team | With AI Team |
|----------------|-------------|
| One model does everything | 4 specialized AI experts collaborate |
| No architecture phase | ChatGPT designs the system first |
| No code review | Gemini reviews every output |
| Hallucinated APIs & outdated docs | Perplexity researches real-time info |
| Manual copy-paste between tabs | Automatic orchestration |
| You merge everything yourself | Claude synthesizes the final result |
| Needs expensive API keys | Uses your existing browser subscriptions |
---
## Features
- **4 AI agents** working together on every task
- **Browser cookie auth** — uses your existing subscriptions (ChatGPT Plus, Gemini Advanced, Perplexity Pro)
- **No API keys required** — zero extra cost
- **Parallel execution** — agents work simultaneously when possible
- **Pipeline mode** — Research → Architecture → Code → Review
- **MCP compatible** — works natively inside Claude Code
- **CLI included** — also works standalone from any terminal
- **Your keys stay local** — credentials never leave your machine
- **62 tests** — full smoke test coverage for all agents
- **CI pipeline** — automated linting and testing
---
## Authentication Setup
Each AI uses **your own subscription** via browser cookies. No middleman, no extra costs.
| Service | What You Need | Cost |
|---------|--------------|------|
| ChatGPT | Session token from browser | Free (your Plus/Pro sub) |
| Gemini | `__Secure-1PSID` cookie from browser | Free (your Advanced sub) |
| Perplexity | Cookie string from browser | Free (your Pro sub) |
| Claude | Already logged into Claude Code | Free (your Claude sub) |
---
### ChatGPT Setup
1. Log into [chatgpt.com](https://chatgpt.com) in your browser
2. Open this URL in the same browser: `https://chatgpt.com/api/auth/session`
3. Copy the `accessToken` value (the long string)
4. In Claude Code: `Use ai_team_login with service='chatgpt', token='<paste>'`
> Token expires after a few weeks. Just repeat these steps to refresh.
---
### Gemini Setup
1. Log into [gemini.google.com](https://gemini.google.com) in your browser
2. Open DevTools (`F12`) → **Application** → **Cookies** → `gemini.google.com`
3. Find `__Secure-1PSID` — copy its value
4. *(Optional)* Also copy `__Secure-1PSIDTS` for longer sessions
5. In Claude Code: `Use ai_team_login with service='gemini', token='<paste 1PSID value>'`
> Uses the `gemini-webapi` library under the hood. Your Gemini Advanced subscription gives you access to the latest models.
**Alternative:** You can also use a free API key from [Google AI Studio](https://aistudio.google.com/apikey) (500 req/day).
---
### Perplexity Setup
1. Log into [perplexity.ai](https://perplexity.ai) in your browser
2. Open DevTools (`F12`) → **Network** tab
3. Click any request to `perplexity.ai`
4. Find the **Cookie** request header → copy the entire value
5. In Claude Code: `Use ai_team_login with service='perplexity', token='<paste>'`
> Uses your Pro/Max subscription. Cookies expire when you log out.
---
### Claude Setup
Claude uses your existing Claude Code CLI login. If you're already using Claude Code, you're done.
```bash
# Only if not logged in yet:
claude login
```
---
### Verify All Agents
```bash
python3 aiteam.py status
```
```
AI TEAM STATUS
========================================
Claude Code READY
ChatGPT Plus READY
Gemini Advanced READY
Perplexity Pro READY
```
---
## Usage
### From Claude Code (MCP) — Recommended
Just talk naturally:
```
> Ask ChatGPT to design the database schema for a todo app
> Have Gemini review this file for security issues
> Get Perplexity to research best practices for Redis caching
> Run the full AI team on implementing OAuth2
```
#### Team Mode
Say **`aiteam start`** in Claude Code to activate team mode. Every response becomes a team effort — Claude automatically consults all agents before answering.
Say **`aiteam stop`** to return to normal.
### CLI Commands
| Command | What it does |
|---------|-------------|
| `python3 aiteam.py team "task"` | Full pipeline: research → design → code → review |
| `python3 aiteam.py plan "task"` | ChatGPT designs architecture |
| `python3 aiteam.py review "task"` | Gemini reviews for bugs & security |
| `python3 aiteam.py research "task"` | Perplexity researches latest info |
| `python3 aiteam.py code "task"` | Claude implements |
| `python3 aiteam.py status` | Check which agents are ready |
| `python3 aiteam.py login <service>` | Login to a service |
#### Global CLI (optional)
```bash
cd ai-team-mcp/ai-team
chmod +x aiteam.sh
ln -s "$(pwd)/aiteam.sh" /usr/local/bin/aiteam
# Now use from anywhere:
aiteam team "Build a REST API with auth"
```
---
## Architecture
```
You (Terminal / Claude Code)
|
AI Team MCP Server
|
+----------+--------+--------+-----------+
| | | |
Claude ChatGPT Gemini Perplexity
(Coder) (Architect) (Reviewer) (Researcher)
| | | |
+----------+--------+--------+-----------+
|
Combined Result
```
### MCP Tools
| Tool | Description |
|------|-------------|
| `ai_team_status` | Check which agents are online and ready |
| `ai_team_login` | Save login credentials for a service |
| `ask_chatgpt` | Send a task to ChatGPT (architecture/planning) |
| `ask_gemini` | Send a task to Gemini (code review/security) |
| `ask_perplexity` | Send a task to Perplexity (research/docs) |
| `ai_team_run` | Run the full 4-agent pipeline |
| `ai_team_chat` | Collaborate with ChatGPT in team mode |
---
## Examples
```bash
# Full team pipeline
aiteam team "Build a REST API with authentication using FastAPI"
# Individual agents
aiteam research "Best Python ORMs in 2026"
aiteam plan "Microservice architecture for e-commerce"
aiteam review "Check my auth middleware for security vulnerabilities"
aiteam code "Implement WebSocket real-time notifications"
```
### Try These Prompts in Claude Code
```
"Build a Stripe payment integration"
"Design a scalable notification system"
"Review this codebase for SQL injection vulnerabilities"
"Research the latest changes in React 19"
"Compare Redis vs Memcached for session storage"
"Build a CI/CD pipeline for this project"
```
---
## Project Structure
```
ai-team-mcp/
├── ai-team/
│ ├── agents/
│ │ ├── base.py # Base agent class
│ │ ├── claude_agent.py # Claude Code integration
│ │ ├── chatgpt_agent.py # ChatGPT (session token)
│ │ ├── gemini_agent.py # Gemini (gemini-webapi + API key)
│ │ ├── perplexity_agent.py # Perplexity (cookies + API key)
│ │ ├── session_store.py # Secure credential storage
│ │ └── team.py # Team coordinator & pipeline
│ ├── tests/
│ │ ├── test_agents.py # 43 agent smoke tests
│ │ ├── test_prime.py # Demo module tests
│ │ └── conftest.py # Test configuration
│ ├── mcp_server.py # MCP server (7 tools)
│ ├── aiteam.py # CLI interface
│ ├── aiteam.sh # Global shell command
│ ├── requirements.txt # Python dependencies
│ └── sessions.example.json # Credentials template
├── .github/workflows/ci.yml # CI pipeline (ruff + pytest)
├── CLAUDE.md # Team mode instructions
└── README.md
```
---
## Security
- **Your keys stay local.** Credentials stored in `sessions.json` on your machine only.
- **`sessions.json` is gitignored.** Never committed to the repo.
- **No data sent to us.** All API calls go directly to AI providers.
- **No middleman.** No proxy server, no third-party relay.
- **Open source.** Audit every line of code yourself.
---
## Troubleshooting
| Problem | Fix |
|---------|-----|
| ChatGPT "Unauthorized" | Get fresh token from `chatgpt.com/api/auth/session` |
| Gemini not responding | Re-copy `__Secure-1PSID` cookie from browser |
| Perplexity 401 error | Re-copy cookies from DevTools → Network tab |
| MCP not found | Run `claude mcp add ai-team -s user python3 /path/to/mcp_server.py` |
| `ModuleNotFoundError` | Run `pip install -r requirements.txt` |
| Agent "NOT LOGGED IN" | Run login step for that service |
---
## Roadmap
- [x] Claude Code integration (MCP)
- [x] ChatGPT Plus (browser session token)
- [x] Gemini Advanced (browser cookies via gemini-webapi)
- [x] Perplexity Pro (browser cookies)
- [x] Full team pipeline (research → design → code → review)
- [x] CLI interface
- [x] Team mode for Claude Code
- [x] 62 automated tests + CI pipeline
- [ ] Grok integration
- [ ] DeepSeek integration
- [ ] Local LLM support (Ollama, LM Studio)
- [ ] Web dashboard
- [ ] VS Code extension
- [ ] Conversation memory
- [ ] Streaming responses
---
## Contributing
Contributions are welcome! See [CONTRIBUTING.md](ai-team/CONTRIBUTING.md) for details.
### Good First Issues
- New AI provider integrations (Grok, DeepSeek, Mistral, Ollama)
- Better orchestration strategies
- VS Code / JetBrains extensions
- Web dashboard
- Streaming responses
---
## Disclaimer
This project uses **unofficial browser cookie authentication** for ChatGPT, Gemini, and Perplexity. These methods rely on reverse-engineered internal APIs and third-party libraries (`gemini-webapi`, `perplexity`). They may break at any time if providers change their internal endpoints. When they break, update the dependencies — the upstream libraries are actively maintained.
For Gemini, you can alternatively use a **free official API key** from [Google AI Studio](https://aistudio.google.com/apikey) which is stable and supported.
This tool is for **personal use** with your own subscriptions. Your credentials are stored locally and never transmitted to any third party.
---
## License
MIT License. Free to use, modify, and distribute.
---
<p align="center">
<b>Stop asking one AI to do everything.<br>Build an AI engineering team instead.</b>
<br><br>
<a href="https://github.com/needsomehelp/ai-team-mcp">⭐ Star this repo</a> if it saves you time.
<br>
<a href="https://github.com/needsomehelp/ai-team-mcp/issues">Report a bug</a> · <a href="https://github.com/needsomehelp/ai-team-mcp/pulls">Submit a PR</a> · <a href="https://github.com/needsomehelp/ai-team-mcp/discussions">Discussions</a>
</p>
MCP Config
Below is the configuration for this MCP Server. You can copy it directly to Cursor or other MCP clients.
mcp.json
Connection Info
You Might Also Like
No recommendations available