Build production agents, step by step.

Build a working agent loop with no framework: tool calls, memory, and control flow you fully own.

1. 1Build your first agent loop
2. 2Define structured tool schemas
3. 3Add memory and conversation state
4. 4Handle tool errors and retries
5. 5Ship it: package and run your agent

Drop an agent into an app you already ship, without rewriting your codebase or your data model.

1. 1Pick the sidecar vs. inline pattern
2. 2Expose scoped, audited DB tools
3. 3Wire the agent to your auth context
4. 4Deploy the agent alongside your app

Coordinate specialist agents: routing, hand-offs, and shared state across a multi-agent system.

1. 1Design a router/worker topology
2. 2Pass context between agents cleanly
3. 3Add a supervisor and hand-offs
4. 4Parallelize and merge agent results

Measure agents like software: metrics, regression suites, and traces that catch silent failures.

1. 1Pick 5 metrics that catch regressions
2. 2Build a repeatable eval harness
3. 3Trace and debug failing runs