Cost and Latency Budgets for Agents

Course

Build Your Own Agent Harness

Udemy: coming soon

Tools

• LiteLLM (open-source, free)
• Langfuse (open-source, freemium)

Watch first

• Evaluate a Harness: Same Model, Different Harness
• Caching: Prompt Caching and Semantic Caching

In this module

1. The Eight Parts of an Agent Harness
2. How Coding-Agent Harnesses Are Built (Claude Code, Codex)
3. Skills and Plugins: Package Know-How for Agents
4. Hooks: Run Your Own Code Before and After Every Tool Call
5. Human Approvals: Permission Modes and Allow/Deny Rules
6. Context Engineering for Long Tasks: Compaction and Notes
7. Sandboxing Agents: Containers, Files, Network, Secrets
8. Project: Build a Mini Coding-Agent Harness
9. Evaluate a Harness: Same Model, Different Harness
10. Cost and Latency Budgets for Agents