Loop Engineering is a free, no-signup toolkit for designing the system around an AI agent — not just the prompt. Describe a task and it generates copyable /goal prompts for Claude Code & Codex, a runnable continuous-claude command, and a safe GitHub Actions skeleton. It also has a token-cost calculator, a loop readiness score, AGENTS.md/SKILL.md generators, templates, and a failure-case library. Everything runs in your browser — nothing uploaded or executed.

Hey hunters! I kept hand-prompting coding agents turn by turn — and watching them delete tests to make CI green, burn tokens with no progress, or merge broken PRs. So I built Loop Engineering: free, in-browser tools to design the loop around an agent (goal, validation, boundaries, stop rule, budget, feedback) instead of babysitting it. You describe a task and it generates copyable /goal prompts for Claude Code & Codex, a continuous-claude command, and a safe GitHub Actions skeleton — plus a token-cost calculator, a readiness score, AGENTS.md/SKILL.md generators, loop templates, and a "ways loops fail" library. It's free, no sign-up, and runs entirely in your browser — nothing is uploaded or executed. Would love your feedback on the templates and the cost model 🙏

How does the loop readiness score actually work behind the scenes, is it a fixed rubric or does it adapt based on the task you describe?

browser-only and no signup is a nice touch, actually used the failure-case library to tighten up one of my agent loops and it caught a dumb retry loop i hadn't noticed

How does the loop readiness score actually work behind the scenes, is it a fixed rubric or does it adapt based on the task you describe?

browser-only and no signup is a nice touch, actually used the failure-case library to tighten up one of my agent loops and it caught a dumb retry loop i hadn't noticed

Loop Engineering

Design safe, token-aware AI agent loops — free, no sign-up

Design safe, token-aware AI agent loops — free, no sign-up