A practical guide to AI agents in production: event loops, tool contracts, memory tiers, evaluation, deployment flags. Grounded in build-your-own-openclaw and the long-horizon ai-long-task framework.
Inside the ebook
What you get
Architect an event-driven agent that survives multi-day tasks
Design tool contracts that the agent can compose reliably
Tier memory across session, project, and persistent stores
Evaluate agents on task success, safety, cost, and latency
Deploy agents with flags, graduated rollouts, and kill switches
Inside
Why event-driven beats request-response for agents
Tool contracts and typed function calls
Memory tiers and MAP-Elites-style solution archives