Agents Towards Production is an open-source collection of production-grade tutorials for every building block of a GenAI agent stack. Each tutorial lives in its own folder with runnable notebooks or scripts, so you can move from concept to working agent in minutes. The goal is to close the gap between a prototype and a system you can confidently ship. If you are an ML engineer, AI developer, or software engineer moving your first agents from notebook to production, this is where you start.Documentation Index
Fetch the complete documentation index at: https://mintlify.com/NirDiamant/agents-towards-production/llms.txt
Use this file to discover all available pages before exploring further.
Quickstart
Clone the repo and run your first tutorial in under ten minutes.
Architecture
Understand the full production agent stack before diving into tutorials.
What you will learn
The tutorials cover every horizontal layer of a production agent system. You do not need to work through them in order — each one is self-contained with its own dependencies and README.Agent frameworks
Orchestration and workflow design with LangGraph, FastAPI, MCP, and Kotlin Koog.
Memory & knowledge
Persistent state, semantic search, and RAG with Redis, Mem0, Cognee, and Contextual AI.
Tool integration & data
Secure tool calling, real-time web search, and large-scale web data collection.
Deployment
Docker containers, AWS Bedrock AgentCore, on-premises LLMs with Ollama, and GPU cloud.
Security
Input/output guardrails, prompt injection defenses, and automated security testing.
Observability & evaluation
Tracing, behavioral evaluation, fine-tuning, and multi-agent coordination protocols.
All 22 tutorials
Stateful workflows with LangGraph
Design multi-step agent workflows using a directed graph architecture with persistent state.
Agents as APIs with FastAPI
Create and deploy agents as performant APIs with synchronous and streaming endpoints.
Tool & API integration via MCP
Integrate agents with external tools using the Model Context Protocol.
AI agents in Kotlin with Koog
Build your first agent in Kotlin using JetBrains’ Koog framework, from hello world to tool calling.
Dual-memory & semantic search (Redis)
Implement short-term and long-term memory with semantic search and persistent state.
Self-improving memory with Mem0
Build agents with hybrid vector and graph memory that automatically evolves with each interaction.
AI memory with Cognee
Transform scattered development data into unified knowledge graphs with contextual insights.
Production-ready RAG with Contextual AI
Build enterprise-grade RAG systems with intelligent indexing and automated evaluation.
Secure tool calling with Arcade
Enable agents to call external tools — Gmail, Slack, Notion — with OAuth2 and approval workflows.
Real-time web search with Tavily
Give agents access to live web data for research, monitoring, and up-to-date recommendations.
Web data collection with Bright Data
Collect and process web data at scale with enterprise-grade scraping infrastructure.
Containerizing agents with Docker
Package agents for portability and scalability across environments.
AWS Bedrock AgentCore deployment
Transform local agents into production-ready managed services on AWS Bedrock.
On-premises LLMs with Ollama
Run large language models locally for privacy, cost control, and low-latency workflows.
GPU deployment with RunPod
Deploy agents on scalable GPU infrastructure for demanding compute workloads.
Agent tracing with LangSmith
Add comprehensive observability to capture traces, decision points, and timing data.
Automated evaluation with IntellAgent
Automate behavioral analysis and performance metrics to improve agent quality.
Security with LlamaFirewall
Apply input, output, and tool security guardrails covering prompt injection and behavior alignment.
Security testing with Apex
Run hands-on prompt injection attacks, defenses, and automated security testing.
Fine-tuning for domain expertise
Fine-tune language models for specialized agent behavior, domain knowledge, and cost efficiency.
Multi-agent communication with A2A
Simulate collaborative agent workflows using open communication protocols for interoperability.
Chatbot UI with Streamlit
Build a web-based chatbot with file upload, session state, and a chat interface for agent demos.
How the tutorials are structured
Every tutorial follows the same structure so you can orient yourself quickly:README.md
README.md
A high-level overview of the tutorial: what it covers, why it matters, what you will learn, and a link to the notebook or code.
Notebook or tutorial.md
Notebook or tutorial.md
An interactive Jupyter notebook (
.ipynb) or a markdown guide with screenshots, step-by-step instructions, and runnable code cells.app.py or equivalent
app.py or equivalent
The main application code — well-commented and production-style, so you can adapt it directly to your own project.
requirements.txt
requirements.txt
Pinned dependency versions so the tutorial runs exactly as written.
Each tutorial is completely self-contained. You can clone the repo and jump directly to any tutorial folder without completing the others first.