AI Agent

autonomous-agents | agent-loop

Overview

An AI agent is a system that uses an LLM to plan and execute actions in the real world or digital environments. Agents typically combine a language model with tools, memory, and a loop that observes, plans, and acts.

Key Components

• Planning: Task decomposition, subgoal generation
• Tool use: Code execution, web search, file operations
• Memory: Short-term (context window) and long-term (vector DB)
• Loop architecture: agent-loop handles the observe→plan→act cycle

Relationship to Other Projects

• AutoGPT and babyagi are early reference implementations
• claude-code and agent-code are modern coding-specialized agents
• MetaGPT extends agents with multi-role collaboration

References

• See agent-loop-architecture for detailed loop patterns
• See agent-swarm for multi-agent coordination