agentic tools

Standardized software interfaces that enable LLMs to interact with external APIs, execute code, and manipulate real-world data autonomously.

Agentic tools bridge the gap between static reasoning and active execution by providing models with a defined set of capabilities. Developers utilize frameworks like LangChain or CrewAI to connect agents to specific services: querying live data via Serper, running secure code in E2B sandboxes, or updating records in Salesforce. By leveraging structured schemas (often Pydantic or JSON), models like GPT-4o and Claude 3.5 Sonnet can determine the correct tool and parameters for a task with high precision. This architecture enables complex workflows: an agent might autonomously triage a GitHub repository, pull relevant documentation, and submit a pull request without human intervention.

https://python.langchain.com/docs/concepts/tools/