🤖 Autonomous Learning Agent

Designing an Autonomous Learning Agent with Checkpoint Verification and Feynman Pedagogy

A comprehensive deep research system built from scratch using LangGraph, implementing multi-agent coordination, iterative research, and intelligent report generation. The system follows a three-phase architecture: Scope → Research → Write.

---

✨ Features

• 🔍 User Clarification — Determines if additional context is needed before research begins
• 📋 Brief Generation — Transforms conversations into structured research questions
• 🔎 Iterative Research — Agent-based research with custom tools and search integration
• 🌐 MCP Integration — Model Context Protocol for standardized tool access
• 👥 Multi-Agent Supervisor — Coordinates parallel research agents for complex topics
• 📝 Report Generation — Synthesizes findings into comprehensive reports
• 🐳 Docker Ready — Full Docker + docker-compose setup with LangGraph Studio

---

📚 Tutorial Notebooks

This repo contains 6 progressive tutorial notebooks building a deep research system:

#	Notebook	Focus
1	1_scoping.ipynb	User clarification & brief generation
2	2_research_agent.ipynb	Research agent with custom tools
3	3_research_agent_mcp.ipynb	Research agent with MCP servers
4	4_research_supervisor.ipynb	Multi-agent supervisor coordination
5	5_full_agent.ipynb	Complete end-to-end system
6	6_checkpoint_agent.ipynb	Checkpoint verification agent

---

🛠️ Tech Stack

Component	Technology
Framework	LangGraph, LangChain
LLM Providers	Groq, Google Gemini, OpenAI, Anthropic
Search	Tavily API
Protocol	Model Context Protocol (MCP)
Package Manager	uv
Containerization	Docker + docker-compose
Monitoring	LangSmith (optional)
Code Quality	ruff, mypy

---

🚀 Quickstart

Option 1: Docker (Recommended)

# Clone the repository
git clone https://github.com/your-username/Autonomous-Learning-Agent.git
cd Autonomous-Learning-Agent

# Copy and configure environment variables
cp .env.example .env
# Edit .env with your API keys

# Build and run
docker-compose build
docker-compose up -d

Access the services:

Service	URL
LangGraph Studio	https://smith.langchain.com/studio/?baseUrl=http://127.0.0.1:8000
API Documentation	http://127.0.0.1:8000/docs

Option 2: Local Installation

# Prerequisites: Python 3.11+, Node.js, uv
uv sync
cp .env.example .env
# Edit .env with your API keys

# Run LangGraph server
uvx --refresh --from "langgraph-cli[inmem]" --with-editable . --python 3.11 langgraph dev --allow-blocking

# Or run notebooks
uv run jupyter notebook

---

🔑 Environment Variables

Create a .env file from .env.example:

TAVILY_API_KEY=your_tavily_api_key       # Required: Web search
GROQ_API_KEY=your_groq_api_key           # Required: LLM inference
GOOGLE_API_KEY=your_google_api_key       # Optional: Gemini models
LANGSMITH_API_KEY=your_langsmith_key     # Optional: Tracing
LANGSMITH_TRACING=true
LANGSMITH_PROJECT=deep_research_from_scratch

⚠️ Never commit your .env file. Use .env.example as a template.

---

🏗️ Architecture

User Query → Scope Phase → Research Phase → Write Phase → Final Report
                 │               │
           Clarification    Multi-Agent
           Brief Gen.       Supervisor
                            ├── Agent 1
                            ├── Agent 2
                            └── Agent N

Three-Phase Workflow

1. Scope — Clarify research scope and generate structured research briefs
2. Research — Iterative research using agents with tools (Tavily, MCP servers)
3. Write — Synthesize all findings into a comprehensive report

---

📁 Project Structure

Autonomous-Learning-Agent/
├── notebooks/                          # Tutorial notebooks (SOURCE OF TRUTH)
│   ├── 1_scoping.ipynb                 # User clarification & brief generation
│   ├── 2_research_agent.ipynb          # Research agent with custom tools
│   ├── 3_research_agent_mcp.ipynb      # MCP server integration
│   ├── 4_research_supervisor.ipynb     # Multi-agent supervisor
│   ├── 5_full_agent.ipynb              # Complete end-to-end system
│   ├── 6_checkpoint_agent.ipynb        # Checkpoint verification
│   └── utils.py                        # Shared utilities
├── src/deep_research_from_scratch/     # Generated source code (DO NOT EDIT)
│   ├── autonomous_learning_agent.py
│   ├── multi_agent_supervisor.py
│   ├── research_agent.py
│   ├── prompts.py
│   ├── state_*.py
│   └── utils.py
├── Dockerfile
├── docker-compose.yml
├── langgraph.json
├── pyproject.toml
├── .env.example
└── README.md

Important: The notebooks in notebooks/ are the source of truth. Source code in src/ is auto-generated from notebooks using %%writefile magic. See CLAUDE.md for development guidelines.

---

🎯 Key Learning Outcomes

• Structured Output — Pydantic schemas for reliable AI decision making
• Async Orchestration — Parallel coordination vs synchronous simplicity
• Agent Patterns — ReAct loops, supervisor patterns, multi-agent coordination
• Search Integration — External APIs, MCP servers, content processing
• State Management — Complex state flows across subgraphs

---

📜 License

This project is open source and available under the MIT License.

---

🙋‍♀️ Author

Sujitha Kotyada — @sujitha-kotyada