Poe o3 MCP Server
About
A lightweight MCP server implementation for accessing OpenAI's o3 model via Poe API
Details
- Author
- Anansitrading
- Downloads
- 298
- Categories
- Other
Jump to
- Simple MCP server implementation using FastMCP
- Direct integration with Poe’s API for o3 and other models
- Model selection via command‑line style flags in prompts
- Asynchronous request handling for efficient processing
- Comprehensive error handling and logging
- Easy setup and configuration
Setting up with Highlight
This MCP is not yet compatible with Highlight’s one-click setup. However, you can still use it with Highlight by following these steps:
- Download and install Highlight from highlightai.com/download
- Navigate to the plugins tab and select "Add Custom Plugin"
-
Configure the plugin with the settings below
Plugin Name
Poe o3 MCP ServerCommand (node, npx, python, etc.)Please refer to the README for specific instructions on how to obtain API keys or other required environment variables.
- Enable "Start Automatically" if you want the plugin to start when Highlight launches
From the repository
Install prerequisites (Python 3.8+ and a Poe API key), clone the repository, create a virtual environment, install dependencies, set the POE_API_KEY in a .env file, then run python poe_o3_mcp_server.py. The server listens on standard input/output and exposes two tools: o3_query (sends queries to Poe models, defaulting to “o3”) and ping (returns “pong”). Model selection is done by adding a command‑line‑style flag to the prompt (e.g., --Claude-3.5-Sonnet).
Claude Desktop / Cursor
Paste into your MCP client config file to install this server.
{
"mcpServers": {
"poe o3 mcp server": {
"po3_MCP": {
"command": "python",
"args": [
"-m",
"venv",
"venv"
]
}
}
}
}McpServers
{
"po3_MCP": {
"command": "python",
"args": [
"-m",
"venv",
"venv"
]
}
}Poe o3 MCP Server
A lightweight Model Context Protocol (MCP) server implementation that provides access to OpenAI's o3 model and other models via Poe's API. This server allows you to integrate Poe's AI capabilities into any MCP-compatible application.
Features
- Simple MCP server implementation using FastMCP
- Direct integration with Poe's API to access the o3 model and other models
- Model selection via command-line style flags in prompts
- Asynchronous request handling for efficient processing
- Comprehensive error handling and logging
- Easy setup and configuration
Prerequisites
- Python 3.8+
- A Poe API key (obtainable from https://poe.com/api_key)
Installation
1. Clone this repository:
git clone https://github.com/Anansitrading/po3_MCP.git
cd po3_MCP
2. Create a virtual environment (recommended):
python -m venv venv
source venv/bin/activate # On Windows: venv\Scripts\activate
3. Install the required dependencies:
pip install -r requirements.txt
4. Set up your environment variables:
cp sample.env .env
5. Edit the
.env file and add your Poe API key: POE_API_KEY=your_poe_api_key_here
Usage
Running the MCP Server
Run the server with:
python poe_o3_mcp_server.py
The server will start and listen for MCP protocol messages on standard input/output.
Model Selection via Flags
You can select different models available on Poe by adding a flag to your prompt:
--Claude-3.5-Sonnet Tell me about quantum computing
This will route your query to the Claude-3.5-Sonnet model instead of the default o3 model.
The flag can be placed anywhere in the message:
- At the beginning: --GPT-4 What is the capital of France?
- In the middle: Tell me --Claude-3-Opus about the history of Rome
- At the end: What are the three laws of robotics? --Claude-3.5-Sonnet
The flag will be automatically removed from the message before it's sent to the model.
If no flag is specified, the server defaults to using the "o3" model.
Integrating with MCP Clients
This server provides two tools:
1. o3_query - Send a query to the o3 model (or another model via flags) and get a response
2. ping - A simple test tool that returns "pong"
Example of using the server with an MCP client:
```python
from mcp.client import MCPClient
Sign in to leave a review
Use Google, GitHub, or an email account so ratings stay tied to real people.
No reviews posted yet.
