Perplexity Chat MCP Server

The Perplexity MCP Server provides a Python-based interface to the Perplexity API, offering tools for querying responses, maintaining chat history, and managing conversations. It supports model configuration via environment variables and stores chat data locally. Built with Python and setuptools, it's designed for integration with development environments.

The MCP Server is desined to mimick how users interact with the Perplexity Chat on their browser by allowing your models to ask questions, continue conversations, and list all your chats.

<a href="https://glama.ai/mcp/servers/0nggjl0ohi"> <img width="380" height="200" src="https://glama.ai/mcp/servers/0nggjl0ohi/badge" /> </a>

Components

Tools

• ask_perplexity: Request expert programming assistance through Perplexity. Focuses on coding solutions, error debugging, and technical explanations. Returns responses with source citations and alternative suggestions.
• chat_perplexity: Maintains ongoing conversations with Perplexity AI. Creates new chats or continues existing ones with full history context. Returns chat ID for future continuation.
• list_chats_perplexity: Lists all available chat conversations with Perplexity AI. Returns chat IDs, titles, and creation dates (displayed in relative time format, e.g., "5 minutes ago", "2 days ago"). Results are paginated with 50 chats per page.
• read_chat_perplexity: Retrieves the complete conversation history for a specific chat. Returns the full chat history with all messages and their timestamps. No API calls are made to Perplexity - this only reads from local storage.

Key Features

• Model Configuration via Environment Variable: Allows you to specify the Perplexity model using the PERPLEXITY_MODEL environment variable for flexible model selection.

  You can also specify PERPLEXITY_MODEL_ASK and PERPLEXITY_MODEL_CHAT to use different models for the ask_perplexity and chat_perplexity tools, respectively.

  These will override PERPLEXITY_MODEL. You can check which models are available on the Perplexity documentation.

• Persistent Chat History: The chat_perplexity tool maintains ongoing conversations with Perplexity AI. Creates new chats or continues existing ones with full history context. Returns chat ID for future continuation.

• Streaming Responses with Progress Reporting: Uses progress reporting to prevent timeouts on slow responses.

Quickstart

Prerequisites

Before using this MCP server, ensure you have:

• Python 3.10 or higher
• uvx package manager installed

Note: Installation instructions for uvx are available here.

Configuration for All Clients

To use this MCP server, configure your client with these settings (configuration method varies by client):

"mcpServers": {
  "mcp-perplexity": {
    "command": "uvx",
    "args": ["mcp-perplexity"],
    "env": {
      "PERPLEXITY_API_KEY": "your-api-key",
      "PERPLEXITY_MODEL": "sonar-pro",
      "DB_PATH": "chats.db"
    }
  }
}

Environment Variables

Configure the MCP Perplexity server using the following environment variables:

Variable	Description	Default Value	Required
PERPLEXITY_API_KEY	Your Perplexity API key	None	Yes
PERPLEXITY_MODEL	Default model for interactions	sonar-pro	No
PERPLEXITY_MODEL_ASK	Specific model for ask_perplexity tool	Uses PERPLEXITY_MODEL	No
PERPLEXITY_MODEL_CHAT	Specific model for chat_perplexity tool	Uses PERPLEXITY_MODEL	No
DB_PATH	Path to store chat history database	chats.db	No
WEB_UI_ENABLED	Enable or disable web UI	false	No
WEB_UI_PORT	Port for web UI	8050	No
WEB_UI_HOST	Host for web UI	127.0.0.1	No
DEBUG_LOGS	Enable detailed logging	false	No

Using Smithery CLI

npx -y @smithery/cli@latest run @daniel-lxs/mcp-perplexity --config "{\"perplexityApiKey\":\"pplx-abc\",\"perplexityModel\":\"sonar-pro\"}"

Usage

ask_perplexity

The ask_perplexity tool is used for specific questions, this tool doesn't maintain a chat history, every request is a new chat.

The tool will return a response from Perplexity AI using the PERPLEXITY_MODEL_ASK model if specified, otherwise it will use the PERPLEXITY_MODEL model.

chat_perplexity

The chat_perplexity tool is used for ongoing conversations, this tool maintains a chat history. A chat is identified by a chat ID, this ID is returned by the tool when a new chat is created. Chat IDs look like this: wild-horse-12.

This tool is useful for debugging, research, and any other task that requires a chat history.

The tool will return a response from Perplexity AI using the PERPLEXITY_MODEL_CHAT model if specified, otherwise it will use the PERPLEXITY_MODEL model.

list_chats_perplexity

Lists all available chat conversations. It returns a paginated list of chats, showing the chat ID, title, and creation time (in relative format). You can specify the page number using the page argument (defaults to 1, with 50 chats per page).

read_chat_perplexity

Retrieves the complete conversation history for a given chat_id. This tool returns all messages in the chat, including timestamps and roles (user or assistant). This tool does not make any API calls to Perplexity; it only reads from the local database.

Web UI

The MCP Perplexity server now includes a web interface for easier interaction and management of chats.

Features

• Interactive chat interface
• Chat history management
• Real-time message display

Screenshots

Chat List View

Chat Interface

Accessing the Web UI

When WEB_UI_ENABLED is set to true, the web UI will be available at http://WEB_UI_HOST:WEB_UI_PORT.

By default, this is http://127.0.0.1:8050.

Development

This project uses setuptools for development and builds. To get started:

1. Create a virtual environment:

   python -m venv .venv
   source .venv/bin/activate  # On Linux/macOS
   # or
   .venv\Scripts\activate  # On Windows
   

2. Install the project in editable mode with all dependencies:

   pip install -e .
   

3. Build the project:

   python -m build
   

The virtual environment will contain all required dependencies for development.

Contributing

This project is open to contributions. Please see the CONTRIBUTING.md file for more information.

License

This project is licensed under the MIT License. See the LICENSE file for details.