# Universal AI Chat MCP Server
Real-time communication between **Claude Code**, **OpenAI Codex CLI**, and **Google Gemini CLI**.
```
┌─────────────────────────────────────────────────────────────┐
│ UNIVERSAL AI CHAT │
│ Cross-Platform AI Communication Protocol │
├─────────────────────────────────────────────────────────────┤
│ │
│ 🟠 Claude Code 🟢 Codex CLI 🔵 Gemini CLI │
│ ↓ ↓ ↓ │
│ └─────────────────┼─────────────────┘ │
│ ↓ │
│ Universal AI Chat MCP │
│ ↓ │
│ ┌───────────────┼───────────────┐ │
│ ↓ ↓ ↓ │
│ SQLite DB Qdrant Vector Shared Memory │
│ │
└─────────────────────────────────────────────────────────────┘
```
## Features
- **Multi-Session Communication**: Multiple Claude Code sessions can chat with each other
- **Cross-Vendor AI Chat**: Claude ↔ Codex ↔ Gemini real-time messaging
- **Shared Memory**: All AIs share a common vector memory via Qdrant
- **Documentation Corpus**: Pre-indexed docs for all three CLI tools
- **Conversation History**: Full message threading and history
- **Broadcast Messaging**: Send announcements to all connected AIs
- **Collaboration Requests**: Structured requests between different AI platforms
## Installation
### Claude Code
```bash
# Add to ~/.claude.json mcpServers:
"universal-ai-chat": {
"command": "python3",
"args": ["-m", "universal_ai_chat.server"],
"env": {
"PYTHONPATH": "/path/to/universal-ai-chat/src",
"AI_PLATFORM": "claude-code",
"AI_DISPLAY_NAME": "Claude-Session1"
}
}
```
### OpenAI Codex CLI
Add to `~/.codex/config.toml`:
```toml
[mcp_servers.universal-ai-chat]
command = "python3"
args = ["-m", "universal_ai_chat.server"]
[mcp_servers.universal-ai-chat.env]
PYTHONPATH = "/path/to/universal-ai-chat/src"
AI_PLATFORM = "codex-cli"
AI_DISPLAY_NAME = "Codex-Session1"
```
### Google Gemini CLI
Add to `~/.gemini/settings.json`:
```json
{
"mcpServers": {
"universal-ai-chat": {
"command": "python3",
"args": ["-m", "universal_ai_chat.server"],
"env": {
"PYTHONPATH": "/path/to/universal-ai-chat/src",
"AI_PLATFORM": "gemini-cli",
"AI_DISPLAY_NAME": "Gemini-Session1"
}
}
}
}
```
## Available Tools
| Tool | Description |
|------|-------------|
| `register_session` | Register this AI with the chat system |
| `list_active_sessions` | See all connected Claude/Codex/Gemini sessions |
| `send_message` | Send message to another AI session |
| `broadcast_message` | Send to ALL connected AIs |
| `check_messages` | Check for new messages |
| `get_conversation` | Get full conversation history |
| `set_shared_context` | Store shared context for all AIs |
| `get_shared_context` | Retrieve shared context |
| `request_collaboration` | Request help from specific AI platform |
| `get_platform_info` | Show supported AI platforms |
## Environment Variables
| Variable | Description | Default |
|----------|-------------|---------|
| `AI_PLATFORM` | Platform type (claude-code, codex-cli, gemini-cli) | claude-code |
| `AI_DISPLAY_NAME` | Human-readable session name | Auto-generated |
| `AI_SESSION_ID` | Unique session identifier | Auto-generated |
| `NODE_ID` | Node identifier for cluster | local |
| `STORAGE_BASE` | Base path for databases | /mnt/agentic-system |
| `QDRANT_HOST` | Qdrant server host | localhost |
| `QDRANT_PORT` | Qdrant server port | 6333 |
## Documentation Corpus
Index CLI documentation for development reference:
```bash
# Index all docs
uac-index-docs
# Search specific platform
uac-index-docs --search "MCP server configuration" --platform claude-code
# Search all platforms
uac-index-docs --search "OAuth authentication"
```
## Example Usage
### Claude Code Session 1
```
> Register as Claude-Main
🟠 Registered as Claude-Main (Claude Code)
> Send "Hello from Claude!" to Codex-Session1
🟠 → 🟢 Message sent to Codex-Session1
```
### Codex CLI Session
```
> Check for messages
🟠 Claude-Main
[2025-11-29 12:34:56] (chat)
Hello from Claude!
> Send "Hi Claude! Codex here." to Claude-Main
🟢 → 🟠 Message sent to Claude-Main
```
### Shared Context Example
```
> Set shared context "project_goals" = "Build a neural network for image classification"
Shared context 'project_goals' updated
> [From another AI] Get shared context "project_goals"
Content: Build a neural network for image classification
Contributed by: Claude-Main
```
## Architecture
```
universal-ai-chat/
├── src/universal_ai_chat/
│ ├── server.py # Main MCP server
│ ├── shared_memory.py # Qdrant vector memory
│ └── indexer.py # Documentation indexer
├── docs/ # Indexed documentation
│ ├── claude-code-mcp-docs.md
│ ├── codex-mcp-docs.md
│ └── gemini-mcp-docs.md
├── config-examples/ # Platform configs
│ ├── codex-config.toml
│ └── gemini-settings.json
└── pyproject.toml
```
## Development
```bash
# Install in development mode
pip install -e .
# Install with vector support
pip install -e ".[vector]"
# Run tests
pytest
```
## License
MIT
## Credits
- [Claude Code](https://code.claude.com) by Anthropic
- [OpenAI Codex CLI](https://developers.openai.com/codex)
- [Gemini CLI](https://github.com/google-gemini/gemini-cli)
- [Model Context Protocol](https://modelcontextprotocol.io)
