Skip to content

Getting Started

Get ResearchCrew running in 5 minutes and conduct your first research.

Prerequisites

  • Python 3.10+ (< 3.14)
  • UV package manager
  • API Keys for your LLM provider (OpenRouter, OpenAI, Anthropic, etc.)

Installation

1. Install UV

pip install uv

UV is a fast, reliable Python package manager. If you prefer pip, that works too.

2. Clone and Install Dependencies

git clone https://github.com/yourusername/researchcrew.git
cd researchcrew
crewai install

This installs CrewAI and all dependencies listed in pyproject.toml.

3. Configure Environment

cp .env.example .env

Edit .env and add your API keys. Example:

OPENROUTER_API_KEY=sk-your-key-here
OPENROUTER_MODEL_NAME=openouter/openai/gpt-4o-mini

Which LLM should I use?

  • Fast research: Use Haiku or GPT-4o-mini (cheaper, faster)
  • Accurate research: Use GPT-4, Claude 3 Opus (more reliable citations)

See Configuration for more options.

4. Run Your First Research

crewai run

This runs the crew with the default input. You'll see:

✓ Research Planner completed
✓ Web Crawler completed
✓ Content Extractor completed
✓ Synthesis Researcher completed
✓ Reporting Analyst completed

Report saved to: outputs/<yyyymmdd>.md # in the format yyyymmdd.md

Open <yyyymmdd>.md to see your research findings with citations!

File Structure

After running, you'll see these output files:

researchcrew/
├── inputs/input.md
└── outputs/<yyyymmdd>.md

Provide Your First Research Topic

Create input.md in the root folder:

# Research Topic: The Future of AI in Healthcare

Research what are the most significant developments in AI-assisted medical diagnosis in 2024-2025. 
Include emerging trends, specific tools, and expert perspectives.

Then run:

crewai run

The crew will:

  1. Plan the research strategy
  2. Search for authoritative sources
  3. Extract verified claims with sources
  4. Synthesize findings into coherent insights
  5. Generate a cited report

Iterative Research (Next Round)

After reviewing <yyyymmdd>.md, add your feedback to the end of the report:

## User Feedback

Please explore more on:
- Real-world implementation challenges in hospitals
- Cost-benefit analysis of AI diagnostic tools

Skip further research on:
- Historical background of AI (already covered)

Then run again:

crewai run

The crew will see your feedback, remember previous research, and iterate accordingly. The next report will be saved as outputs/<yyyymmdd>.md.

Next Steps

  • Architecture — Understand the 5-agent pipeline
  • Usage Guide — Single-round and multi-round workflows
  • Features — Learn about reliability checks and citations
  • Examples — See complete research walkthroughs

Troubleshooting

"ModuleNotFoundError: No module named 'crewai'"

Reinstall dependencies:

crewai install

"API Key is invalid"

Check your .env file and verify your API key is correct for your chosen provider.

"No output directory found"

Create an output/ folder in the root:

mkdir output

"Crew seems stuck or running slowly"

  • Check your internet connection (web crawler needs it)
  • Your LLM provider may be rate-limited; try again in a few minutes
  • Check your API key quotas/credits