No-Code and Low-Code Platforms
- OpenAI Agent Mode: An emerging feature in ChatGPT that allows you to give a goal and let the AI break it into subtasks, use built-in tools, and return results more autonomously.
- n8n: A powerful open-source automation tool where AI can be integrated into workflows for triggering actions and connecting data sources.
- Zapier + AI: Combines Zapier’s vast integration library with AI models to trigger intelligent actions without coding.
Developer-Focused Frameworks
- LangChain: A Python and JavaScript framework for building AI agents that can reason, take actions, and use memory across tasks. Highly flexible and well-documented.
- Haystack: An open-source framework for search, retrieval, and question-answering agents.
- Auto-GPT Variants: Experimental open-source agents that can run multi-step goals, useful for rapid prototyping and testing agentic behaviors.
Infrastructure and Support Tools
- Vector Databases: Tools like Pinecone, Weaviate, and FAISS for storing and retrieving semantic embeddings — essential for agents with long-term memory.
- APIs and Data Sources: Public datasets, custom APIs, and knowledge bases that feed agents with the information they need to reason and act effectively.
- Model Providers: Platforms like OpenAI, Anthropic, and local deployment tools such as Ollama for running models privately.
From Exploration to Building
In this series, I’ve been focused on exploring the “what” and “why” of agentic AI. In my next series, Building Agentic AI, I’ll shift to the “how” — walking through the process of selecting tools, designing workflows, and testing agents in real scenarios, including for my SDG ecosystem project.
Coming Next
In the first post of the Building Agentic AI series, I’ll compare no-code, low-code, and full-code options, helping you decide which path to take based on your skills, goals, and resources.