Hi Product Hunt 👋 I'm Taha, the maker of Pitroom.

I built it because my Claude Code sessions kept burning premium tokens on reading: grepping, opening files, scanning code they never quoted. The expensive model was doing work a cheap one could do.

Pitroom gives your agent a pit crew. Claude Code, Codex or any MCP agent hands the legwork (research, fixes, tests, reviews) to cheaper workers running in parallel: OpenCode with free models, Codex, Gemini CLI, Qwen Code, Kilo Code and more. Your agent keeps the decisions and gets back only a verified answer, the exact diff and a cost receipt.

A few things I cared about:

• Code changes happen in an isolated copy; nothing lands until your agent reviews it

• File references in answers are checked, not trusted

• A live dashboard (pitroom dash) shows every worker as it runs

• On eight questions about a 419k-line repo, workers read 5.5M tokens and handed the agent 7,684 (715× less)

It's free and open source (MIT): github.com/ATASTECH/pitroom

I'd love to hear which tasks you'd hand off first, and where it falls short. I'll be around all day answering questions 🙏