Deep Research Agent

A production-grade multi-agent research system built with LangChain DeepAgents.

Overview

This project implements a sophisticated research agent that can:

• Plan: Break down complex research tasks into manageable steps
• Search: Query the web for relevant information
• Validate: Cross-check and verify findings
• Synthesize: Generate comprehensive research reports

Built on the DeepAgents framework, which implements patterns from production agents like Claude Code and Manus.

Architecture

┌─────────────────────────────────────────────────────────────┐
│                    RESEARCH ORCHESTRATOR                     │
│                    (Main DeepAgent)                          │
├─────────────────────────────────────────────────────────────┤
│                                                             │
│  ┌─────────────┐  ┌─────────────┐  ┌─────────────┐         │
│  │   Planner   │  │  Researcher │  │   Writer    │         │
│  │  Sub-Agent  │  │  Sub-Agent  │  │  Sub-Agent  │         │
│  └─────────────┘  └─────────────┘  └─────────────┘         │
│                                                             │
├─────────────────────────────────────────────────────────────┤
│  Built-in Tools: todos, filesystem, task delegation         │
│  Custom Tools: web_search                                   │
├─────────────────────────────────────────────────────────────┤
│  Middleware: Planning, Summarization, HITL, Caching         │
└─────────────────────────────────────────────────────────────┘

Features

• DeepAgents Architecture: Planning, sub-agents, filesystem, detailed prompts
• Multi-Model Support: Claude (Anthropic) and GPT (OpenAI)
• Web Search: Tavily integration for AI-optimized search
• Production Ready: Proper logging, configuration, error handling
• LangSmith Integration: Tracing and observability

Quick Start

Prerequisites

• Python 3.11 or higher
• API keys for:
  • Anthropic (Claude) OR OpenAI (GPT)
  • Tavily (web search)
  • LangSmith (optional, for tracing)

Installation

# Clone the repository
git clone https://github.com/yourusername/deep-research-agent.git
cd deep-research-agent

# Create virtual environment
python -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate

# Install dependencies
pip install -e .

# For development
pip install -e ".[dev]"

Configuration

# Copy example environment file
cp .env.example .env

# Edit .env with your API keys
# Required: ANTHROPIC_API_KEY or OPENAI_API_KEY
# Required: TAVILY_API_KEY
# Optional: LANGSMITH_API_KEY

Usage

# Run the research agent
python main.py

# Or with a specific topic
python main.py "Your research topic here"

Project Structure

deep-research-agent/
├── src/
│   ├── agents/          # DeepAgent definitions
│   ├── tools/           # Custom tools (web search, etc.)
│   ├── config/          # Configuration management
│   └── utils/           # Utilities (logging, etc.)
├── tests/               # Test files
├── docs/                # Documentation
├── pyproject.toml       # Dependencies
├── .env.example         # Environment template
└── main.py              # Entry point

Learning Path

This project is designed as a learning journey through DeepAgents:

1. Phase 1: Basic agent with web search
2. Phase 2: Planning and task management
3. Phase 3: Multi-agent orchestration
4. Phase 4: Quality and iteration loops
5. Phase 5: Human-in-the-loop and persistence
6. Phase 6: Report generation

Built With

• DeepAgents - Agent harness
• LangGraph - Agent runtime
• LangChain - LLM framework
• Tavily - AI search API

License

MIT License - see LICENSE for details.

Acknowledgments

• LangChain team for DeepAgents and LangGraph
• Anthropic for Claude Code (architectural inspiration)
• The AI agent community