AI Agents · TRACK 02/05
Tool Use & Tool Design
Giving agents hands: designing tools, controlling calls, and handling failures.
// THE TRACK
01 · START HERETool Use & Function CallsLearn how a text-only model gains the ability to search, run code, and call APIs — and what actually happens during a tool call.BEGINNERDesigning Tool SchemasUnderstand the schema design principles, naming conventions, error message patterns, and testing strategies that make LLM agent tools reliable.INTERMEDIATEWeb Search & BrowsingCompare search APIs, page-fetch tools, and full browser control as ways to get live web data into your agent.INTERMEDIATECode Execution vs More ToolsSee why letting an agent write and run code often outperforms a sprawling tool catalog — and when a dedicated tool still wins.INTERMEDIATENames, Descriptions & GranularityLearn the naming, description, and scoping rules that decide whether your agent picks the right tool or flails.INTERMEDIATEHandling Tool ErrorsLearn how to turn tool failures into signals the model can recover from — clear error messages, safe retries, and guards against retry loops.INTERMEDIATEParallel Tool CallsUnderstand how agents fire multiple tool calls in a single turn, when that genuinely speeds things up, and the dependency traps it creates.INTERMEDIATEStructured Tool OutputsLearn that designing what a tool returns matters as much as its inputs — and how trimmed, structured results make agents more accurate and cheaper.INTERMEDIATEAG-UI ProtocolYou will understand what the AG-UI protocol standardizes, how its event stream connects an agent backend to a user-facing app, and how it complements protocols like MCP and A2A.INTERMEDIATE