autoresearch

name	autoresearch
description	Automated research pipeline — find, index, and surface insights from papers, docs, and web sources.
when_to_use	When asked to research a topic deeply, find papers, build a knowledge base, or create a research report with citations.
allowed_tools	["web_search","web_fetch","firecrawl","memory_search","file_write","file_read"]
context	inline

AutoResearch Pipeline

How to Research Deeply

Step 1: Decompose the Query

Break the research question into 3-5 sub-queries:

• Core concept definition
• Recent developments (last 6 months)
• Key papers/implementations
• Practical applications
• Open problems

Step 2: Multi-Source Search

For each sub-query:

1. web_search — find relevant pages and papers
2. web_fetch — read the most promising results
3. firecrawl — if a source has multiple relevant pages, crawl the site
4. Save key findings to memory with memory_search context

Step 3: Synthesize

• Cross-reference findings across sources
• Identify consensus vs. disagreement
• Note recency — prefer 2025-2026 sources
• Cite everything: [1] source.com - Title

Step 4: Index for Future Use

Save the research summary to agent memory so it can be retrieved later:

• Key facts as explicit memories
• Patterns as inductive memories
• Hypotheses as abductive memories

Research Quality Checklist

• Multiple sources consulted (minimum 3)
• Recent sources prioritized
• Conflicting viewpoints noted
• All claims cited
• Summary saved to memory for future reference

Paper Discovery

When looking for research papers:

1. Search arxiv.org, semanticscholar.org, paperswithcode.com
2. Use firecrawl to extract paper abstracts and key findings
3. Index papers with: title, authors, date, key contribution, relevance score