Hi Product Hunt! I’m Cassio, the maker of ForgeLoop. AI coding tools can produce changes quickly. Understanding what changed, which checks actually ran, and whether a task is complete takes more than a confident final message. ForgeLoop gives that work an explicit record: • Define the task’s contract and working boundaries. • Follow verification gates before proceeding. • Run checks with command provenance and execution receipts. • Resume existing work across sessions. • Validate completion against current evidence. Your coding agent implements the changes. ForgeLoop manages the protocol and checks the completion record. The optional ForgeLoopAudit companion provides a read-only view of tasks, evidence and findings. Stable 2.0.0 passed 16/16 fresh-user checks on each native platform: Windows, Linux and macOS. The reports and acceptance boundaries are public. The core requires Node.js 24.19.0+. Semantic checkpoints require TypeSafe/Jev access. ForgeLoop is free for noncommercial use under its included license; Audit remains an unsigned preview. I’d love feedback from developers using AI coding agents: what evidence do you need before you trust that a task is finished?

ForgeLoop

Contracts, checks and evidence for AI coding workflows

Contracts, checks and evidence for AI coding workflows