Polymarket MCP Tool
# Polymarket MCP Tool
A Model Context Protocol (MCP) tool for interacting with Polymarket through Claude Desktop.
## Prerequisites
- Claude Desktop App for Mac
- Python 3.8+
## Installation
1. Clone this repository:
```bash
git clone https://github.com/fernandezpablo85/polymarket-mcp.git
cd polymarket-mcp
```
2. Install `uv` if you haven't already:
```bash
curl -LsSf https://astral.sh/uv/install.sh | sh
```
3. Install dependencies:
```bash
uv sync
```
## Configure Claude Desktop
1. Open Claude Desktop configuration directory:
```bash
open ~/Library/Application\ Support/Claude
```
2. Create or edit `claude_desktop_config.json`:
```bash
touch ~/Library/Application\ Support/Claude/claude_desktop_config.json
```
3. Add the following configuration:
```json
{
"mcpServers": {
"polymarket": {
"command": "/Users/YOUR_USERNAME/.local/bin/uv",
"args": [
"--directory",
"/Users/YOUR_USERNAME/projects/polymarket-mcp",
"run",
"main.py"
]
}
}
}
```
Replace `YOUR_USERNAME` with your actual macOS username.
## Usage
Invoke the tool via Claude Desktop.
## Troubleshooting
- If tools don't appear in Claude Desktop:
- Verify your `claude_desktop_config.json` is correct
- Restart Claude Desktop
- Check your Python path and dependencies
- If authentication fails:
- Verify your `.env` file has correct credentials
- Check Polymarket API status
## License
MIT
## Contributing
Feel free to open issues or submit pull requests.
TDQS
Scored across 3 tools
The three tools have distinct primary purposes: getting trending topics, fetching markets by topic, and searching markets by term. However, there is some functional overlap between `prediction_markets_markets_by_topic` and `search_prediction_markets` as both return market data, which could cause minor confusion about when to use each.
All tool names follow a consistent snake_case pattern with a clear `prediction_markets_` prefix and descriptive suffixes (`trending_topics`, `markets_by_topic`, `search_prediction_markets`). The naming is predictable and follows a logical structure throughout.
With only 3 tools, the server feels somewhat thin for a prediction markets domain. While the tools cover basic discovery and search functionality, the limited count suggests potential gaps in comprehensive market interaction capabilities that might be expected from such a service.
The tools provide good discovery and search capabilities but lack any market interaction functions (like placing bets, viewing user positions, or managing portfolios). For a prediction markets server, this represents notable gaps in the full lifecycle of market participation, though the provided tools work well together for information gathering.