MCP Perplexity Server

What is MCP Perplexity Server?

MCP Perplexity Server provides an API to ask questions to a model behind the Perplexity API, wrapped in the Model Context Protocol (MCP). It is intended for developers who want to integrate Perplexity AI’s conversational search into MCP‑compatible applications.

How to use MCP Perplexity Server?

Set up the server via Docker (docker compose up) or Node.js (pnpm install && pnpm start after setting environment variables). You must configure MCP_PERPLEXITY_API_KEY with your Perplexity API key. The server runs on stdio (default) or SSE transport; for SSE, the endpoint is http://localhost:8080/sse by default.

Key features of MCP Perplexity Server

• Provides MCP‑compatible API for Perplexity AI models
• Supports both stdio and SSE transport protocols
• Configurable default model and model list
• Optional bearer authentication for SSE endpoints
• Easy deployment via Docker Compose or Node.js

Use cases of MCP Perplexity Server

• Integrate Perplexity’s conversational search into MCP‑based AI assistants
• Build a question‑answering tool that uses Perplexity’s up‑to‑date web knowledge
• Run a local MCP server that proxies queries to Perplexity for MCP clients

FAQ from MCP Perplexity Server

Do I need a Perplexity API key?

Yes, the MCP_PERPLEXITY_API_KEY environment variable is required and must contain your Perplexity API key.

What transport methods are supported?

The server supports stdio (default) and SSE transport, configured by the MCP_TRANSPORT variable.

How can I run the server?

You can run it using Docker Compose with the provided docker-compose.yml or directly via Node.js with pnpm install && pnpm start.

What models are available?

The default model is sonar-pro; you can set a comma‑separated list of models via MCP_PERPLEXITY_MODELS (default: sonar,sonar-pro).

Is authentication required for SSE?

No, but you can enable optional bearer authentication by setting MCP_SSE_AUTH_HEADER_VALUE to a token.