Agentic AI Runs on Tools
2 days ago
- #Agentic AI
- #LLMs
- #Function Calling
- Agentic AI is defined by its ability to have goals, planning, multi-step reasoning, memory/persistence, and interaction with the outside world.
- Agenticness is a spectrum; systems like coding agents are highly agentic, while simple chatbots are less so.
- LLMs without external tools face limitations like information cut-offs and lack of real-time data access.
- Function calling is a key paradigm for LLMs to interact with external tools, enabling more agentic behavior.
- Frameworks like MCP and A2A build on function calling to standardize tool discovery and usage for AI agents.
- Agentic AI is expected to grow as trust in AI tools increases and LLM function-calling accuracy improves.