Ask HN: What's your current AI agentic setup?
Summary
Ask HN collects a snapshot of how readers configure AI agent setups, including subscriptions, local models, and orchestration tools. The thread features real-world mixes of Claude, Codex, Opus, OpenCode, Pi, Hermes, OpenRouter, and self-hosted GLM, with debates on when to self-host, how to orchestrate agents, and how to balance cost, latency, and control. Several commenters describe pipelines that spawn multiple agents, use subagents, and build bespoke harnesses, while others prefer simpler setups.