OpenAlex Explorer

An agent skill, LLM-powered automated literature review pipeline: search papers via OpenAlex, screen for relevance & quality, extract summaries, and write structured reviews.

How It Works

After registering this skill, you simply tell your AI agent what you want to research, and it handles the rest — searching academic papers, screening for relevance and quality, extracting summaries, and writing a structured review.

The pipeline:

1. Keyword Generation — AI brainstorms search keywords from your question
2. Adaptive Search — Searches OpenAlex with async parallelism
3. Relevance & Quality Screening — Two-stage LLM filtering
4. Structured Summaries — Extracts key findings
5. Review Writing — Topic clustering → parallel chapter writing → citation validation

You can also just build a literature database without writing a summaries Note that this skill DOESNOT read the full paper, so it's only suitable for exploration, but this feature also make it token-efficient and fast

Install

Download this skill,then just tell your AI agent something like:

"Read SKILL.md,Install this Skill"

The agent will read the skill definition, ask you for any missing details, configure the JSON spec,then you can use it.

Manual CLI Usage (Optional)

You can also run the pipeline directly:

# Build the paper database
python3 build_db.py specs/my_review.json

# Write the review
python3 write_review.py output/My_Review_20260708_120000/

# Manage sessions
python3 manage.py list
python3 manage.py export 0

Configuration Reference

llm_config.json

Field	Required	Description
model	✅	provider/model-name (e.g., deepseek/deepseek-chat, openai/gpt-4o)
api_key	✅	API credential
temperature	❌	Default 0.3
max_tokens	❌	Default 4096

openalex_config.json

Field	Required	Description
mailto	❌	Email for polite pool (~100 req/s)
api_key	❌	Free at openalex.org

Spec JSON (Research Question)

Field	Required	Default	Description
research_question	✅	—	Your research topic
start_year / end_year	❌	0	Year range
min_papers	❌	15	Target papers
method_preference	❌	""	Preferred methods
research_field	❌	""	Domain restriction
initial_depth	❌	20	Search depth per keyword
max_total_papers	❌	1000	Hard database limit
max_quality_papers	❌	200	Quality-passed limit
relevance_workers	❌	20	Concurrency
quality_workers	❌	20	Concurrency
language	❌	"en"	"en" or "zh-hans"
lang_instruction	❌	""	Extra prompt prefix

Project Structure

├── llm_config.json            ← LLM config (gitignored)
├── openalex_config.json       ← OpenAlex config (gitignored)
├── specs/*.json               ← Research specs (created by agent)
├── SKILL.md                   ← Skill definition
├── build_db.py                ← Database builder
├── write_review.py            ← Review writer
├── manage.py                  ← Session manager
├── src/                       ← Core modules
└── output/                    ← Results (gitignored)

Output

Output

Requirements

• Python >= 3.12
• An LLM API key
• OpenAlex API KEY,it's free

License

MIT