Loading...
A comprehensive glossary of key concepts, tools, and practices in the world of AI agents
A design paradigm where AI operates with autonomy, memory, reasoning, and goals.
The logic an agent uses to plan and execute multi-step actions toward a goal.
The amount of information an LLM can "remember" and reason about during a task.
Agents select and use tools on-the-fly like browsers, APIs, or databases.
The external system or sandbox where the agent observes, acts, and learns.
Agent development kits like LangChain and CrewAI used to build agent systems.
Agents operate with clear objectives and adapt actions to reach their goals.
A method where humans guide, monitor, or correct agent decisions.
How multiple agents collaborate or share tasks using protocols like A2A or OAP.
A core capability for agents to interpret structured inputs and API responses.
Storage of facts or documents agents use for reasoning, including vector or graph DBs.
Stores persistent knowledge for reuse across multiple tasks or sessions.
A standard for feeding memory and tools into LLMs in agent workflows.
Graph-style logic where agents follow modular steps and decision branches (e.g. LangGraph).
Platforms to monitor, debug, and track agent performance (e.g., LangFuse, Helicone).
The craft of writing structured prompts to instruct LLM agents effectively.
How agents understand and break down user intent into executable actions.
A reasoning method where agents reflect on past outputs to self-improve.
The ability to track and update agent memory, task progress, and environment state.
Breaking complex problems into simpler subtasks using reasoning frameworks.
Scoring systems used by agents to evaluate outcomes and select optimal paths.
A database that stores and retrieves embeddings for semantic memory and search.
Coordinating multi-step agent processes using platforms like n8n or LangFlow.
Sometimes needed when agents interact with legacy systems or APIs using XML formats.
Used to define agent roles, tools, memory types, and task flows in many frameworks.
When agents solve problems without prior examples, relying purely on their LLM logic.
These concepts form the foundation of modern agentic AI systems. As the field evolves, new patterns, tools, and methodologies continue to emerge, pushing the boundaries of what autonomous AI agents can achieve.
Explore the fundamental building blocks of modern artificial intelligence, organized like the periodic table of elements.
View AI Periodic Table