Versions:
JYYCode, published by jyycode and currently at version 1.2.6 with two versions available in the catalog, is a multi-agent engineering workflow tool in the AI-assisted software development category that plans, delegates, reviews, revises, and delivers engineering work from a single request. Where conventional AI coding tools rely on one chat box and one agent—forcing users to watch progress step by step, restart from scratch when something goes wrong, and lose context, accountability, and momentum on large tasks—JYYCode upgrades a one-line goal into an organized engineering run that returns reviewed, auditable output. Its operation is enforced by protocol rather than goodwill: planning, dispatch, reporting, and review are governed by a runtime protocol under which child agents can only report and cannot rewrite the plan, while any rejection must state the exact gap, with that feedback automatically carried into the next dispatch. The system is backed by state rather than memory, since plans persist as revisioned structured files with optimistic-concurrency writes, and sessions, snapshots, and the shared blackboard all live in SQLite running in WAL mode, allowing runs to resume precisely after restarts or channel switches. Execution is powered by a team rather than a lone agent: one root agent orchestrates up to twenty parallel sub-agents with distinct roles and models, coordinating over the shared blackboard and presenting structured human-in-the-loop questions at decision points. Typical use cases include delegating complex coding tasks to specialized sub-agents, maintaining continuity across long or interrupted engineering sessions, enforcing accountable review cycles with explicit rejection feedback, and keeping human oversight at critical decisions while the orchestrated team handles implementation, making it suited to developers who need structured, resumable, and auditable AI-driven engineering rather than ad hoc conversational assistance.
Tags: