Skip to content

ResearchCrew Documentation

Welcome to ResearchCrew — an AI-powered research system built for integrity, guidance, and transparency.

What is ResearchCrew?

ResearchCrew is a multi-agent research automation platform that:

  • Conducts guided research — Humans steer the investigation after each round, exploring specific topics in depth
  • Maintains citation integrity — Every claim is traceable to a source URL; no hallucinations
  • Validates reliability — Web crawler filters unreliable sources, and the pipeline includes checks at each stage
  • Remembers context — Long-term memory avoids redundant searches and preserves research history

Unlike generic AI summarization tools, ResearchCrew treats research as an iterative, human-guided process where you actively steer the investigation and trust the output.

Core Workflow

1. User provides research topic
        ↓
2. Crew runs 5-stage pipeline
   (Planner → Crawler → Extractor → Synthesis → Reporter)
        ↓
3. User reviews report and findings
        ↓
4. User provides guidance (explore X deeper, ignore Y)
        ↓
5. Crew iterates with feedback context
        ↓
6. Refined, cited research output

Key Features

Human-in-the-Loop Research

After each iteration, you review the findings and actively guide the next round. The system learns your feedback and refines accordingly. This isn't a one-shot summarization—it's a collaborative investigation.

Citation Integrity

Every claim in the final report is backed by a source URL. The crew extracts verbatim quotes from pages, preserves source context through synthesis, and outputs publication-ready citations.

Multi-Stage Reliability

  • Web Crawler applies quality filters (excludes Medium, Reddit, unreliable domains)
  • Content Extractor records confidence levels for each claim
  • Synthesis flags contradictions between sources
  • Reporter notes "Insufficient data" where gaps exist

Persistent Memory

The system uses LanceDB to store and learn from past research. This enables:

  • Deduplication across runs (avoid re-crawling the same sources)
  • Context preservation (remember what was already researched)
  • Multi-day workflows (continue research across sessions)

Get Started

Examples

Want to see it in action?

Philosophy

Research integrity over speed. Every finding must be verifiable. Every source must be reliable. Every iteration should be guided by human judgment.

ResearchCrew is for researchers, analysts, and teams who need transparent, well-sourced research outputs.