The agent loop

One step of an agent loop Practice: one step of the loop The loop in code Practice: the loop in code Stopping and retrying Practice: stopping and retrying What only code can decide Practice: what code decides

Reliability in production

Timeout and idempotency key Practice: timeout and idempotency key Says done, but nothing happened Practice: false success Durable agent: checkpoints and resume Practice: durable agent

Agent designs

Workflow or agent Practice: workflow or agent ReAct and plan-and-execute Practice: ReAct and plan-and-execute Plan-and-Solve Practice: Plan-and-Solve Reflexion Practice: Reflexion One agent or a swarm Practice: one agent or a swarm

Context and cost

Why the agent bill grows Practice: tokens and cache Context engineering Practice: managing the window Agent memory Practice: what the agent keeps

Tools and protocols

MCP Practice: MCP Tool descriptions and injection Practice: tool descriptions and injection Code instead of a toolset Practice: code instead of a toolset Agent skills and A2A Practice: agent skills and A2A

Frameworks

LangGraph or the raw loop Practice: LangGraph or the raw loop Agent SDKs: what you are buying Practice: agent SDKs

Coding agents

Driving a coding agent from a spec Practice: driving a coding agent from a spec

←→RU Sign in

Agents › Tools and protocols

Practice: MCP

Builds onMCP

Sign in to do the tasks.

‹ Previous Next lesson ›

Suggest an improvementFound a mistake