Forge, don't vibe.

A portable AI engineering lifecycle. Drop it into any repo — your agent drives it and leaves durable artifacts behind, not chat smoke.

Get started

Why loop engineering

One command, then step back

npx agentsmyth init drops a staging folder. Your agent inspects the repo, resolves what setup couldn't fill in, and wires the whole workflow into place.

Seven gates, no skipped steps

Think, Plan, Build, Review, Test, Ship, Reflect. Each gate leaves a readable artifact. Later agents restore context from the record, not memory.

Yours from the first minute

Setup's resolution pass shapes the workflow around your domain, your paths, your release rules. Nothing is mandatory unless your config says so.

Evidence or it did not happen

Validators refuse silent drift. Commands need output. Skipped checks are recorded as risk, never hidden.

What you actually lose

Most AI coding sessions end the same way. The code exists, the reasoning behind it evaporated the moment the chat scrolled off screen, and the next agent that touches the repo starts cold.

agentsmyth treats that evaporation as the actual problem. Every phase forces the reasoning, the evidence, and the decisions onto disk, where the next request, the next agent, and the next month can read them. Nothing important lives only in the chat, because the chat is the thing that disappears.

It is not a framework, not a linter, not an agent. It is a contract that the agent you already use agrees to follow.