Agents & Tool Use

Agentic loops, function calling, MCP, evals, multi-agent patterns.

A system where an LLM autonomously plans, calls tools, observes results, and iterates toward a goal rather than producing one-shot output.

e.g. An agent that books a flight by searching, comparing, and confirming.

The repeating cycle of think → act → observe that lets an agent make progress on a task across multiple steps.

e.g. Search, read the result, decide the next search, repeat.

A structured test suite used to measure how well a model or agent performs a task, often used to catch regressions before shipping changes.

e.g. Running 200 saved tasks against a new prompt to check the pass rate.

A model's ability to emit a structured request to invoke an external function or API, then incorporate the result into its response.

e.g. Calling a get_weather(city) function instead of guessing.

An open standard for connecting AI models to external tools, data sources, and systems through a common interface.

e.g. One MCP server exposing both a database and a filesystem to any compatible client.

Multiple specialized agents (or agent instances) coordinating, delegating, or debating to solve a task none could solve as well alone.

e.g. A "planner" agent delegating research to a "researcher" agent.