Basic Workflow
Conduct a single-round research investigation.
Overview
A basic research workflow is a single end-to-end pipeline run:
- Provide topic → Create
inputs/input.mdwith research question - Run crew →
crewai run - Get report → Read
outputs/<yyyymmdd>.mdwith citations
This is ideal for one-off research tasks or exploring a topic for the first time.
Step-by-Step Example
1. Create Your Research Topic
Create input.md in the project root:
# Research Topic: Quantum Computing in 2025
Research the current state of quantum computing in 2025. Include:
- Recent breakthroughs and milestones
- Current limitations and challenges
- Commercial applications and timelines
- Leading companies and organizations
2. Run the Crew
You'll see progress output:
Starting crew execution...
✓ Research Planner: Created research plan
✓ Web Crawler: Found 12 relevant sources
✓ Content Extractor: Extracted 45 claims
✓ Synthesis Researcher: Synthesized findings
✓ Reporting Analyst: Generated report
Output files created:
- outputs/<yyyymmdd>.md (final report)
3. Review the Output
Open outputs/<yyyymmdd>.md:
# Research Report: Quantum Computing in 2025
## Executive Summary
Quantum computing has reached a significant inflection point in 2025, with
[multiple companies reporting quantum advantage in specific applications](https://...).
## Current Breakthroughs
### Error Correction
[IBM announced a breakthrough in error correction](https://...), reducing error rates
by 50% compared to 2024. [Google's Willow chip demonstrated 10-qubit error correction](https://...).
### Commercial Applications
[Financial institutions are testing quantum algorithms for portfolio optimization](https://...),
with [early pilots showing 30% performance improvements](https://...).
## Current Limitations
However, [quantum computers still require extreme cooling to near absolute zero](https://...),
making widespread deployment challenging. [Estimated timeline for practical quantum advantage: 3-5 years](https://...).
## Leading Organizations
- [IBM Quantum](https://...)
- [Google Quantum AI](https://...)
- [IonQ](https://...)
- [Rigetti Computing](https://...)
Every claim includes a citation text you can click.
Output Files
[yyyymmdd].md
Your final research output. This is what you share with others.
Format:
- Markdown with claim citations
- Organized by subtopic
- Notes "Insufficient data" where gaps exist
- Publication-ready quality
Common Patterns
Shallow First Pass
If you want a quick overview:
# Quick Overview: Machine Learning Trends
Provide a 5-minute overview of machine learning trends in 2025,
focusing on recent news and headlines rather than deep analysis.
Run normally. The report will be shorter and faster.
Deep Dive on One Topic
# Research Topic: LLM Safety and Alignment
Focus deeply on AI safety, alignment research, and current approaches to making large language models safer.
Include perspectives from leading researchers and organizations.
The crew will produce a more detailed report.
Comparison Research
# Comparison: Vue vs React in 2025
Compare Vue.js and React frameworks. Include:
- Performance characteristics
- Developer experience and learning curve
- Ecosystem and library support
- Production use cases and adoption
- Strengths and weaknesses of each
The crew will synthesize comparisons across sources.
When to Use Single-Round Research
-
Good for:
-
Initial topic exploration
- Answering a specific question
- Quick competitive analysis
-
First-time research on a topic
-
Not ideal for:
-
Complex, multi-faceted topics (use iterative research)
- Topics requiring human guidance (use human-in-the-loop)
- Research where you want to steer direction (see Iterative Research)
Next Steps
- Iterative Research — Learn how to iterate with feedback
- Configuration — Customize LLM, tools, and outputs
- Architecture — Understand the 5-agent pipeline