Claude Deep Research Template

A production-ready template for conducting systematic, comprehensive research using Claude Code. This implementation leverages an organized flow of data from previous outputs into persistent context, enabling iterative and cumulative research workflows.

What is This?

This template provides a complete, structured workflow for using Claude Code as a deep research assistant. Instead of ad-hoc conversations, this approach:

Systematizes research with clear phases and workflows
Documents everything in organized directories
Generates reusable outputs in multiple formats
Builds on previous work through structured context
Scales from quick investigations to comprehensive studies

Background and Concept

Planning and Development Notes: Claude-Deep-Research-Model Repository
This repository contains my planning and development notes for the concept behind this implementation, which attempts to leverage an organized flow of data from previous outputs into persistent context, among other features.

Concept Discussion Podcast: Listen on Spotify
Audio discussion explaining the concept in detail.

Quick Start

1. Use This Template

Click "Use this template" on GitHub or clone locally:

git clone https://github.com/danielrosehill/Claude-Deep-Research-Template.git my-research-project
cd my-research-project

2. Initialize Your Research

Open the repository in Claude Code and run:

/initiate-research

Claude will interview you about your research needs and set up your project.

3. Start Researching

Follow Claude's guidance through the research workflow:

Context gathering
Prompt planning
Research execution
Output synthesis

Directory Structure

├── .claude/
│   ├── commands/          # Slash commands (e.g., /initiate-research)
│   └── agents/            # Specialized research agents
├── context/               # Research context and source materials
│   ├── from-internet/    # Web research, papers, articles
│   ├── from-human/       # Your context and requirements
│   └── from-history/     # Previous conversation histories
├── prompts/              # Research prompts by stage
│   ├── drafting/         # Draft prompts
│   ├── queue/            # Prompts ready to execute
│   └── run/              # Executed prompts
│       ├── initial/      # Starting prompts
│       └── subsequent/   # Follow-up prompts
├── outputs/              # Research outputs
│   ├── individual/       # Single-topic outputs
│   ├── aggregated/       # Synthesized reports
│   └── reformatted/      # Alternative formats (TTS, SSML)
├── pipeline/             # Workflow automation
├── notes/                # Research notes and logs
└── scratchpad/          # Working area for experiments

How It Works

Research Workflow

Context Gathering - Provide background, objectives, and constraints
Prompt Planning - Claude generates research questions systematically
Research Execution - Execute prompts, document findings, generate follow-ups
Output Synthesis - Aggregate findings into comprehensive reports
Iteration - Build on previous research in subsequent sessions

Specialized Agents

The template includes three specialized agents:

Research Coordinator - Orchestrates the overall workflow
Prompt Generator - Creates effective research questions
Research Synthesizer - Aggregates findings into reports

Key Features

Systematic Research

Structured phases from exploration to synthesis
Clear progression through research depth
Built-in quality checkpoints

Comprehensive Documentation

All findings automatically documented
Research log tracks progress
Conversation histories archived

Multiple Output Formats

Markdown reports
PDF documents
TTS-safe text
SSML for voice synthesis

Reusable Context

Build on previous research sessions
Archive and reference past work
Continuous knowledge accumulation

Slash Commands

/initiate-research - Start a new deep research project

Add your own custom commands in .claude/commands/

Use Cases

This template works for:

Academic Research - Literature reviews, topic exploration
Market Research - Industry analysis, competitive intelligence
Technical Investigation - Technology evaluation, architecture research
Strategic Planning - Trend analysis, scenario planning
Content Creation - Research for articles, presentations, reports
Learning - Deep dives into new subjects

Configuration

CLAUDE.md

The CLAUDE.md file contains instructions for Claude on how to operate in this repository. Customize it for your specific needs.

Custom Slash Commands

Create new slash commands in .claude/commands/:

.claude/commands/my-command.md

Custom Agents

Define specialized agents in .claude/agents/:

.claude/agents/my-agent.md

Best Practices

Start with /initiate-research - Let Claude set up your project properly
Provide rich context - The more context, the better the research
Review and refine - Check Claude's research plan before execution
Synthesize regularly - Don't wait until the end to aggregate
Archive conversations - Save conversation histories for future reference

Tips for Deep Research

Be specific about objectives - Clear goals lead to focused research
Iterate on prompts - Refine research questions based on findings
Use the scratchpad - Experiment without cluttering main outputs
Document methodology - Note how you arrived at conclusions
Build incrementally - Research in sessions, build on previous work

Example Workflow

# 1. Clone template
git clone <template-url> my-research

# 2. Open in Claude Code
cd my-research
claude-code .

# 3. Initialize research
/initiate-research

# 4. Provide context when prompted
# - Research topic
# - Objectives
# - Scope and depth
# - Output requirements

# 5. Review generated research plan
# - Approve or refine

# 6. Let Claude execute research
# - Findings documented automatically
# - Follow-up prompts generated
# - Progress tracked

# 7. Generate synthesis
# - Request aggregated report
# - Multiple formats created

# 8. Continue or conclude
# - Archive session
# - Plan next session
# - Or finalize deliverables

Contributing

Suggestions and improvements welcome! This template is designed to evolve based on real-world usage.

Related Projects

Claude Deep Research Model - Concept and model notes
Claude Code Projects Index - More Claude Code projects

License

MIT License - See LICENSE file for details

Author

Daniel Rosehill

Acknowledgments

Based on the Claude Deep Research Model concept. Special thanks to the Claude Code team at Anthropic for creating such a powerful research tool.