pplx

It's an unofficial CLI program to query/chat with the perplexity API.

Installation

Option 1

• Download the latest release from the releases page.
• Install the binary in /usr/local/bin or any other directory in your PATH.

Option 2: With brew

brew tap sgaunet/homebrew-tools
brew install sgaunet/tools/pplx

Usage

Program to interact with the Perplexity API.

        You can use it to chat with the AI or to query it.

Usage:
  pplx [command]

Available Commands:
  chat        chat subcommand is an interactive chat with the Perplexity API
  help        Help about any command
  query       
  version     print version of pplx

Flags:
  -h, --help   help for pplx

Use "pplx [command] --help" for more information about a command.

Shell Completion

pplx supports shell completion for Bash, Zsh, Fish, and PowerShell, providing intelligent auto-completion for commands, flags, and values.

Features

• Command completion: Auto-complete available commands and subcommands
• Flag completion: Auto-complete flag names with descriptions
• Dynamic value completion: Intelligent completion for:
  • Model names (e.g., sonar, sonar-pro, sonar-deep-research)
  • Search modes (web, academic)
  • Recency filters (hour, day, week, month, year)
  • Context sizes (low, medium, high)
  • Reasoning efforts (low, medium, high)
  • Image formats (jpg, png, gif, etc.)
  • Common domains for search filtering

Quick Installation

The easiest way to set up completions is using the automatic installer:

# Auto-detect your shell and install
pplx completion install

# Or specify a shell explicitly
pplx completion install bash
pplx completion install zsh
pplx completion install fish
pplx completion install powershell

To uninstall:

pplx completion install --uninstall

Manual Installation

If you prefer manual installation, you can generate completion scripts for your shell:

Bash

For current session only:

source <(pplx completion bash)

Permanent installation:

Linux:

pplx completion bash | sudo tee /etc/bash_completion.d/pplx

macOS (with Homebrew):

pplx completion bash > $(brew --prefix)/etc/bash_completion.d/pplx

Zsh

Prerequisites: If shell completion is not already enabled, add this to your ~/.zshrc:

autoload -U compinit; compinit

For current session only:

source <(pplx completion zsh)

Permanent installation:

# Create completions directory if it doesn't exist
mkdir -p ~/.zsh/completions

# Generate completion script
pplx completion zsh > ~/.zsh/completions/_pplx

# Add to your ~/.zshrc (if not already present)
fpath=(~/.zsh/completions $fpath)
autoload -U compinit; compinit

Fish

For current session only:

pplx completion fish | source

Permanent installation:

pplx completion fish > ~/.config/fish/completions/pplx.fish

PowerShell

For current session only:

pplx completion powershell | Out-String | Invoke-Expression

Permanent installation:

# Generate completion script
pplx completion powershell > pplx-completion.ps1

# Add to your PowerShell profile
# Find your profile location with: $PROFILE
# Then add this line to your profile:
. /path/to/pplx-completion.ps1

Using Completions

Once installed, you can use tab completion for commands and flags:

# Complete command names
pplx <TAB>
# Shows: chat, completion, config, help, mcp-stdio, query, version

# Complete flag names
pplx query --m<TAB>
# Completes to: --model

# Complete model names
pplx query --model <TAB>
# Shows: sonar, sonar-pro, sonar-reasoning, sonar-deep-research, etc.

# Complete search modes
pplx query --search-mode <TAB>
# Shows: web, academic

# Complete recency values
pplx query --search-recency <TAB>
# Shows: hour, day, week, month, year

# Complete multiple values for array flags
pplx query --search-domains <TAB>
# Shows: github.com, stackoverflow.com, medium.com, etc.

Troubleshooting

Completions not working after installation:

1. Restart your shell or source your shell configuration file:
   • Bash: source ~/.bashrc or source ~/.bash_profile
   • Zsh: source ~/.zshrc
   • Fish: source ~/.config/fish/config.fish
   • PowerShell: . $PROFILE

Zsh completions not loading:

• Ensure compinit is called in your ~/.zshrc
• Verify the completion file is in your fpath by running echo $fpath
• Try running compinit manually to reload completions

Permission errors during installation:

• Use the automatic installer which handles permissions correctly
• Or use sudo when writing to system directories

Advanced Usage

Save to a custom file:

pplx completion bash -o ~/my-completions/pplx.bash

Generate for a specific shell without installing:

pplx completion zsh > completions.zsh

View all completion subcommands:

pplx completion --help

Chat

Chat with the Perplexity API.

pplx chat

Query

Query the Perplexity API.

pplx query -p "what are the best citations of Jean Marc Jancovici ?" -s "you're a politician"

The above command will return in console a result that looks like:

Query Examples

Basic Queries

# Simple query
pplx query -p "What is the capital of France?"

# Query with system prompt
pplx query -p "Explain quantum computing" -s "You are a physics professor"

# Query with custom model
pplx query -p "Latest AI news" --model "llama-3.1-sonar-large-128k-online"

Advanced Search Options

# Search only from specific domains (using short flag)
pplx query -p "climate change research" -d nature.com,science.org

# Get recent information only (last week) - using short flag
pplx query -p "stock market updates" -r week

# Location-based query
pplx query -p "weather forecast" --location-lat 48.8566 --location-lon 2.3522 --location-country FR

Response Enhancement

# Include images in the response (using short flag)
pplx query -p "Famous landmarks in Paris" -i

# Get related questions (using short flag)
pplx query -p "How to learn programming" -q

# Filter images by format and domain
pplx query -p "Nature photography" -i --image-formats jpg,png --image-domains unsplash.com,pexels.com

Generation Parameters

# Control response length (using short flag)
pplx query -p "Summarize War and Peace" -T 500

# Fine-tune creativity and randomness (using short flags)
pplx query -p "Write a haiku about coding" -t 0.8 --top-p 0.95

# Adjust frequency and presence penalties
pplx query -p "Explain machine learning concepts" --frequency-penalty 0.5 --presence-penalty 0.3

Combined Examples

# Technical research with specific sources and recent data (using short flags)
pplx query -p "Latest developments in quantum computing" \
  -d arxiv.org,nature.com \
  -r month \
  -q \
  -T 1000

# Local business search with images (using short flags)
pplx query -p "Best restaurants near me" \
  --location-lat 40.7128 \
  --location-lon -74.0060 \
  --location-country US \
  -i \
  -r week

# Creative writing with custom parameters (using short flags)
pplx query -p "Write a short story about AI" \
  -s "You are a creative science fiction writer" \
  -t 0.9 \
  -k 50 \
  -T 2000

Available Options

Common Options (for both chat and query)

Option	Short	Type	Description
--model	-m	string	AI model to use
--frequency-penalty		float64	Penalize frequent tokens (0.0-2.0)
--max-tokens	-T	int	Maximum tokens in response
--presence-penalty		float64	Penalize already present tokens (0.0-2.0)
--temperature	-t	float64	Response randomness (0.0-2.0)
--top-k	-k	int	Consider only top K tokens
--top-p		float64	Nucleus sampling threshold
--timeout		duration	HTTP request timeout
--search-domains	-d	[]string	Filter search to specific domains
--search-recency	-r	string	Filter by time: day, week, month, year
--search-mode	-a	string	Search mode: web (default) or academic
--search-context-size	-c	string	Search context size: low, medium, or high
--location-lat		float64	User location latitude
--location-lon		float64	User location longitude
--location-country		string	User location country code
--return-images	-i	bool	Include images in response
--return-related	-q	bool	Include related questions
--stream	-S	bool	Enable streaming responses
--image-domains		[]string	Filter images by domains
--image-formats		[]string	Filter images by formats

Query-specific Options

Option	Short	Type	Description
--user-prompt	-p	string	User question/prompt (required)
--sys-prompt	-s	string	System prompt to set AI behavior

Configuration Files

pplx supports YAML configuration files to manage default settings and create reusable profiles for different use cases. This eliminates the need to specify the same flags repeatedly.

Quick Start

# Initialize a new configuration file
pplx config init

# View current configuration
pplx config show

# Validate configuration
pplx config validate

Configuration File Locations

pplx searches for configuration files in the following order:

1. ./pplx.yaml - Current directory (highest priority)
2. ~/.config/pplx/config.yaml - User config directory
3. /etc/pplx/config.yaml - System-wide configuration (lowest priority)

You can also specify a custom config file:

pplx query -p "your question" --config /path/to/config.yaml

Configuration Precedence

Settings are applied in the following order (later sources override earlier ones):

1. Configuration file defaults
2. Environment variables (e.g., PPLX_API_KEY)
3. Command-line flags (highest priority)

This allows you to set sensible defaults in your config file while still overriding them on the command line when needed.

Configuration Structure

A configuration file has four main sections:

# Default values for all queries
defaults:
  model: sonar
  temperature: 0.2
  max_tokens: 4000

# Search preferences
search:
  mode: web                    # web or academic
  recency: week               # hour, day, week, month, year
  context_size: medium        # low, medium, high
  domains:                    # Optional domain filtering
    - nature.com
    - science.org

# Output preferences
output:
  stream: false
  return_images: false
  return_related: false
  json: false

# API configuration
api:
  timeout: 30s

Environment Variable Interpolation

Configuration values can reference environment variables using ${VAR_NAME} syntax:

api:
  key: ${PPLX_API_KEY}
  timeout: ${PPLX_TIMEOUT:-30s}  # With default fallback

Working with Profiles

Profiles allow you to maintain different configurations for various use cases (research, creative writing, news, etc.).

Creating Profiles

# Create a new profile
pplx config profile create research "Academic research with verified sources"

# List all profiles
pplx config profile list

# Switch active profile
pplx config profile switch research

# Delete a profile
pplx config profile delete creative

Using Profiles in Config Files

# Active profile to use by default
active_profile: research

# Define multiple profiles
profiles:
  research:
    name: research
    description: Academic research with verified sources
    defaults:
      model: llama-3.1-sonar-large-128k-online
      temperature: 0.1
      max_tokens: 2000
    search:
      mode: academic
      context_size: high
      recency: month
      domains:
        - arxiv.org
        - nature.com
        - science.org
    output:
      return_related: true

  creative:
    name: creative
    description: Creative writing and brainstorming
    defaults:
      model: sonar
      temperature: 0.9
      max_tokens: 4000
      frequency_penalty: 0.5
    output:
      stream: true
      return_images: true

  news:
    name: news
    description: Current news and events
    defaults:
      model: sonar
      temperature: 0.2
    search:
      recency: day
      context_size: high
      domains:
        - reuters.com
        - bbc.com
        - apnews.com
    output:
      return_related: true
      return_images: true

Using Profiles from CLI

# Use a specific profile for a query
pplx config profile switch research
pplx query -p "Latest quantum computing research"

# Or override the active profile temporarily
pplx query -p "Write a creative story" --config creative.yaml

Configuration Management Commands

Initialize Configuration

# Create a new config file with defaults
pplx config init

# Create in a specific location
pplx config init --output ~/.config/pplx/config.yaml

# Force overwrite existing config
pplx config init --force

View Configuration

# Show current configuration
pplx config show

# Show specific profile
pplx config show --profile research

# Show configuration from specific file
pplx config show --config /path/to/config.yaml

Validate Configuration

# Validate current configuration
pplx config validate

# Validate specific file
pplx config validate --config /path/to/config.yaml

Edit Configuration

# Open config in default editor
pplx config edit

# Edit specific config file
pplx config edit --config /path/to/config.yaml

Example Use Cases

Research Workflow

Create a research profile for academic queries:

profiles:
  research:
    defaults:
      model: llama-3.1-sonar-large-128k-online
      temperature: 0.1
      max_tokens: 2000
    search:
      mode: academic
      context_size: high
      recency: month
      domains:
        - arxiv.org
        - nature.com
        - science.org
        - ieee.org
    output:
      return_related: true

Use it:

pplx config profile switch research
pplx query -p "Latest breakthroughs in quantum computing"

News Monitoring

Create a news profile for current events:

profiles:
  news:
    defaults:
      temperature: 0.2
      max_tokens: 1500
    search:
      recency: day
      context_size: high
      domains:
        - reuters.com
        - bbc.com
        - apnews.com
        - bloomberg.com
    output:
      return_related: true
      return_images: true

Creative Writing

Create a creative profile with higher temperature:

profiles:
  creative:
    defaults:
      temperature: 0.9
      max_tokens: 4000
      frequency_penalty: 0.5
      presence_penalty: 0.3
    output:
      stream: true
      return_images: true

Example Configuration Files

See the examples/config directory for complete configuration examples:

• default.yaml - General purpose configuration
• research.yaml - Academic research with verified sources
• creative.yaml - Creative writing and brainstorming
• news.yaml - Current news and events

Configuration Best Practices

1. Use profiles for different workflows: Create separate profiles for research, creative work, news monitoring, etc.

2. Set sensible defaults: Configure commonly used options in the defaults section to avoid repetitive flags.

3. Use environment variables for secrets: Store API keys in environment variables rather than directly in config files:

   api:
     key: ${PPLX_API_KEY}

4. Version control your configs: Keep your configuration files in version control (excluding sensitive data) to share setups across machines.

5. Override when needed: Remember that CLI flags always override config file settings, so you can easily adjust behavior for specific queries.

MCP Server (Model Context Protocol)

The pplx mcp-stdio command provides an MCP server that exposes Perplexity AI functionality to Claude Code and other MCP-compatible clients.

Quick Start with Claude Code

# Install the server
brew tap sgaunet/homebrew-tools
brew install sgaunet/tools/pplx
# or download from releases and place in PATH

# Add to Claude Code
claude mcp add perplexity-ai -s user -- pplx mcp-stdio

Manual Configuration

For Claude Code (via config file)

Add to your Claude Code MCP configuration:

{
  "mcpServers": {
    "perplexity-ai": {
      "command": "pplx",
      "args": ["mcp-stdio"],
      "env": {
        "PPLX_API_KEY": "your_perplexity_api_key_here"
      }
    }
  }
}

Configure PPLX_API_KEY with your Perplexity AI API key in your environment or directly in the MCP config.

For Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or equivalent location:

{
  "mcpServers": {
    "perplexity-ai": {
      "command": "/usr/local/bin/pplx",
      "args": ["mcp-stdio"],
      "env": {
        "PPLX_API_KEY": "your_perplexity_api_key_here"
      }
    }
  }
}

For Other MCP Clients

Any MCP-compatible client can use this server by executing:

PPLX_API_KEY=your_key /path/to/pplx mcp-stdio

MCP Tool: query

The MCP server exposes a single powerful tool called query with the following parameters:

Required Parameters

• user_prompt (string): The user question/prompt

Optional Parameters

Core Parameters:

• system_prompt (string): System prompt to guide AI behavior
• model (string): AI model to use (default: sonar-small-online)
• temperature (number): Response randomness (0.0-2.0)
• max_tokens (number): Maximum tokens in response
• frequency_penalty (number): Penalize frequent tokens (0.0-2.0)
• presence_penalty (number): Penalize already present tokens (0.0-2.0)
• top_k (number): Consider only top K tokens
• top_p (number): Nucleus sampling threshold
• timeout (number): HTTP timeout in seconds

Search & Web Options:

• search_domains (array): Filter search to specific domains
• search_recency (string): Filter by time: "day", "week", "month", "year", "hour"
• location_lat (number): User location latitude
• location_lon (number): User location longitude
• location_country (string): User location country code
• search_mode (string): Search mode: "web" or "academic"
• search_context_size (string): Context size: "low", "medium", "high"

Response Enhancement:

• return_images (boolean): Include images in response
• return_related (boolean): Include related questions
• stream (boolean): Enable streaming (collected into complete response)

Image Filtering:

• image_domains (array): Filter images by domains
• image_formats (array): Filter images by formats (jpg, png, etc.)

Response Formats (Sonar models only):

• response_format_json_schema (string): JSON schema for structured output
• response_format_regex (string): Regex pattern for structured output

Date Filtering:

• search_after_date (string): Filter results published after date (MM/DD/YYYY)
• search_before_date (string): Filter results published before date (MM/DD/YYYY)
• last_updated_after (string): Filter results last updated after date (MM/DD/YYYY)
• last_updated_before (string): Filter results last updated before date (MM/DD/YYYY)

Deep Research:

• reasoning_effort (string): For sonar-deep-research model: "low", "medium", "high"

Example Usage in Claude Code

Once configured, you can use the Perplexity MCP server directly in Claude Code:

Search for recent AI developments in computer vision with images

Claude Code will automatically use the MCP server to:

1. Query Perplexity AI with your prompt
2. Filter for recent information
3. Include relevant images
4. Return structured results with citations

Response Format

The MCP server returns JSON with the following structure:

{
  "content": "AI response text with markdown formatting",
  "model": "model_used_for_generation",
  "usage": {
    "prompt_tokens": 123,
    "completion_tokens": 456,
    "total_tokens": 579
  },
  "search_results": [
    {
      "title": "Article Title",
      "url": "https://example.com/article",
      "snippet": "Relevant excerpt..."
    }
  ],
  "images": [
    {
      "url": "https://example.com/image.jpg",
      "description": "Image description"
    }
  ],
  "related_questions": [
    "What are the latest AI breakthroughs?",
    "How is computer vision evolving?"
  ]
}

Environment Variables

• PPLX_API_KEY (required): Your Perplexity AI API key

Troubleshooting

1. Server not starting: Verify PPLX_API_KEY is set
2. Command not found: Ensure pplx is in your PATH
3. Configuration issues: Check JSON syntax in MCP config files
4. API errors: Verify your API key is valid and has sufficient credits

Advanced Configuration Examples

High-quality research with academic sources

{
  "user_prompt": "Latest quantum computing breakthroughs",
  "search_mode": "academic",
  "search_context_size": "high",
  "search_recency": "month",
  "return_related": true,
  "max_tokens": 2000
}

Location-based search with images

{
  "user_prompt": "Best restaurants in Tokyo",
  "location_lat": 35.6762,
  "location_lon": 139.6503,
  "location_country": "JP",
  "return_images": true,
  "image_formats": ["jpg", "png"],
  "search_recency": "week"
}

Structured output for data processing

{
  "user_prompt": "List top 5 programming languages",
  "model": "sonar-small-online",
  "response_format_json_schema": "{\"type\":\"object\",\"properties\":{\"languages\":{\"type\":\"array\",\"items\":{\"type\":\"string\"}}}}",
  "max_tokens": 500
}