Perplexity Search

Use this skill to search the web via the Perplexity Sonar API. It returns accurate, cited, real-time answers grounded in current web sources.

When to use this skill

Invoke this skill automatically when the user asks about:

• Current events, recent news, or anything that may have changed recently
• Product releases, software versions, or API documentation
• Research questions that benefit from citations and web sources
• Comparisons, pricing, or availability that requires live data
• Any question where freshness or sourcing matters

Trigger phrases: "search for", "look up", "find out", "what's the latest", "current", "recent", "news about", "web search", "perplexity search"

Setup

Set your Perplexity API key in the environment before starting OpenClaw:

# Add to ~/.openclaw/.env
PERPLEXITY_API_KEY=pplx-your-key-here

Get a key at https://www.perplexity.ai/settings/api

Command

python3 {baseDir}/scripts/perplexity_search.py --query "<the user's question>"

Optional flags

Flag	Default	Description
--model sonar	sonar	Fast, general-purpose search
--model sonar-pro	—	Higher quality synthesis (costs more)
--model sonar-reasoning	—	Step-by-step reasoning
--domains example.com	—	Comma-separated domain allowlist
--max-results 8	8	Cap displayed search results
--temperature 0.2	0.2	Lower = more factual
--json	—	Print raw JSON response

Usage examples

# General search
python3 {baseDir}/scripts/perplexity_search.py --query "latest AI agent frameworks 2025"

# Higher quality synthesis
python3 {baseDir}/scripts/perplexity_search.py --query "compare Perplexity vs Tavily search APIs" --model sonar-pro

# Restrict to a specific site
python3 {baseDir}/scripts/perplexity_search.py --query "PostgreSQL 17 release notes" --domains postgresql.org

# Reasoning mode for complex topics
python3 {baseDir}/scripts/perplexity_search.py --query "why did SVB fail" --model sonar-reasoning

Output

The script prints a structured markdown report with:

1. Answer — concise, web-grounded response
2. Citations — numbered source URLs
3. Search Results — titles, URLs, and snippets
4. Related Questions — follow-up ideas
5. Usage — token and cost metadata

Agent workflow

1. Identify the user's search intent from their message.
2. Extract a clean, specific query (avoid filler words).
3. Choose sonar for speed, sonar-pro for depth.
4. Run the command with python3 {baseDir}/scripts/perplexity_search.py --query "...".
5. Present the Answer and Citations to the user in a readable format.
6. Offer related questions as follow-up options.

Security notes

• The script reads PERPLEXITY_API_KEY from the environment only — never from arguments.
• The API key is never logged or printed.
• All external calls go only to https://api.perplexity.ai/chat/completions.
• No local files are written.

SECURITY MANIFEST

• Environment variables accessed: PERPLEXITY_API_KEY (only)
• External endpoints called: https://api.perplexity.ai/chat/completions (only)
• Local files read: none
• Local files written: none