Ask HN: What's your current AI agentic setup?
Points and comments are a snapshot, not live.
Ask HN users share their current AI agentic setups, subscriptions, and tools.
The post asks HN users to describe their AI agentic setups, including subscriptions (e.g., Claude, ChatGPT, Cursor, Codex) and tiers, whether they use pay-as-you-go APIs (e.g., OpenRouter) or local models, coding harnesses (e.g., Claude Code, Aider, Pi, OpenCode, custom builds), agent orchestration tools (e.g., Munder Difflin, Gas Town, Ruflo), and broader agent platforms (e.g., OpenClaw, Hermes Agent, Grok bot, Dots, Muse). The author notes their own workflow has changed significantly in the past year and seeks a current snapshot.
What commenters are saying
Many commenters have consolidated to Claude (Code or Desktop) as their primary tool, with some using multiple subscriptions or accounts. Several switched from Cursor or Codex to Claude, citing Opus 5.5's superiority. Some use Claude for code review via sub-agents. A few use custom setups like OpenClaw or Pi with OpenRouter for orchestration, while others avoid agent swarms, preferring to QA outputs. One user runs multiple concurrent agent projects and reports feeling addicted. Others manage costs with DeepSeek Flash API or cheaper tiers.