# Perplexity Research

> Deep Research mode through Perplexity MCP. Use only for broad, important questions where higher cost and longer runtime are acceptable.

- Skill: `nicshik/perplexity-research` (Agent Skill)
- Install (CLI): `npx skillmds@latest add nicshik/perplexity-research`
- Raw SKILL.md: https://api.skillmd.com/api/skills/nicshik/perplexity-research/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: nicshik (https://skillmd.com/u/nicshik)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/nicshik/perplexity-research

---


# Perplexity Research

Use this skill for Deep Research mode through Perplexity MCP.

Observed billing snapshot after skill tests cost `$2.52`; most of it came from `sonar-deep-research`: 478,195 reasoning tokens (`$1.43`), 117 search queries (`$0.59`), 67,825 citation tokens (`$0.14`), and 15,158 output tokens (`$0.12`).

Rules:

- Use only the `perplexity_research` MCP tool from the `perplexity` server.
- Do not silently replace it with `perplexity_search`, `perplexity_ask`, or `perplexity_reason`.
- Keep the question focused before running the tool.
- Return a practical summary with the strongest citations.
- Mention that this mode is slower and more expensive than simple search when relevant.