# Deep Research Perplexity

> Run Perplexity Sonar and Sonar Deep Research for web-grounded answers with citations and bounded search context. Use when a research task needs current web synthesis through Perplexity or comparison against other provider APIs.

- Skill: `closedloop-technologies/deep-research-perplexity` (Agent Skill, multi-file: 5 files)
- Install (CLI): `npx skillmds@latest add closedloop-technologies/deep-research-perplexity`
- Raw SKILL.md: https://api.skillmd.com/api/skills/closedloop-technologies/deep-research-perplexity/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Research & Search
- Author: closedloop-technologies (https://skillmd.com/u/closedloop-technologies)
- Updated: 2026-09-10
- Page: https://skillmd.com/skills/closedloop-technologies/deep-research-perplexity

---


# Perplexity Sonar

Use this skill when the research task needs current web-grounded synthesis and
source citations. The Sonar API is OpenAI-compatible.

## Environment

```bash
op run --env-file .env.climbhill -- python .agents/skills/deep-research-perplexity/scripts/ask.py \
  --prompt "deepresearch the deepresearchers" \
  --model sonar \
  --search-context-size low
```

Required 1Password-backed variable:

- `PERPLEXITY_API_KEY`

## Cost Controls

- Prefer `sonar` or the lowest available Sonar model for smoke tests.
- Use `--search-context-size low` for under-`$1` benchmark smoke tests.
- Use Sonar Deep Research only when the benchmark task truly needs deep
  multi-source reasoning.
- Track token, citation, reasoning, request, and search-query costs when the API
  returns them.

## Current Source Links

- https://docs.perplexity.ai/docs/sonar/quickstart
- https://docs.perplexity.ai/docs/getting-started/pricing