ussi69-dotcom/server-perplexity-ask-wsl2

server-perplexity-ask-wsl2

MCP (Model Context Protocol) server for Perplexity AI API with WSL2 IPv4 compatibility fix.

Why this fork?

The original server-perplexity-ask package uses Node.js native fetch() which has issues with IPv6 on WSL2 (Windows Subsystem for Linux 2). This causes ETIMEDOUT errors even when curl works fine.

This fork fixes the issue by using the https module with family: 4 to force IPv4 connections.

Installation

For Claude Code (MCP)

Add to your ~/.claude.json:

{
  "mcpServers": {
    "perplexity-ask": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "server-perplexity-ask-wsl2"],
      "env": {
        "PERPLEXITY_API_KEY": "your-api-key-here"
      }
    }
  }
}

Manual installation

npm install -g server-perplexity-ask-wsl2

Configuration

Set your Perplexity API key as an environment variable:

export PERPLEXITY_API_KEY="pplx-your-api-key"

Get your API key from: https://www.perplexity.ai/settings/api

Usage

The server provides a single tool perplexity_ask that accepts an array of messages:

{
  "messages": [
    {"role": "user", "content": "What are the latest AI news?"}
  ]
}

The WSL2 Fix

The key change is in how we make HTTP requests:

// Before (broken on WSL2):
fetch(url, options)

// After (works on WSL2):
const https = require('https');
https.request({
  ...options,
  family: 4  // Force IPv4
})

Why does this happen?

1. WSL2 has inconsistent IPv6 connectivity
2. Node.js fetch() tries IPv6 first, waits for timeout, then fails
3. curl works because it tries IPv4/IPv6 in parallel
4. Using family: 4 forces IPv4-only connections

Model

This server uses sonar-pro model by default. See Perplexity API docs for available models.

License

MIT

Credits

• Original package: server-perplexity-ask
• WSL2 fix by: @ussi69-dotcom