Deer Flow

Overview

Deer Flow is a research powerhouse. Unlike simple research agents, it handles long-horizon projects — planning strategies, searching the web, writing and executing analysis code, and producing comprehensive reports with data visualizations. With 57k+ stars, it’s the most popular research agent.

How It Works

1. Plan — Decomposes question into multi-step plan
2. Research — Searches web, reads papers
3. Analyze — Writes and executes Python analysis code
4. Synthesize — Produces report with visualizations
5. Iterate — Refines based on gaps

Use Cases

• Academic research — Literature reviews, data analysis
• Market research — Industry analysis with data
• Technology evaluation — Comparisons with benchmarks
• Investment research — Due diligence with financial analysis

Getting Started

git clone https://github.com/bytedance/deer-flow
cd deer-flow && pip install -r requirements.txt
python main.py --question "Compare top 5 vector databases for RAG"

Example

Question: "Compare top 5 vector databases for production RAG"

Output: 3000-word report with:
- Feature comparison table
- Performance benchmarks (QPS, latency, recall)
- Pricing analysis with charts
- Recommendations by use case

Alternatives

• GPT Researcher — Simpler research agent
• Open Deep Research — Firecrawl-powered research
• Perplexity — AI search (not autonomous)