About Romyq

AI coding assistants are very good at generating code. They are not good at managing a software project. They do not know what phase a project is in, whether tests exist, whether governance rules are being followed, or whether the project is ready to ship.

The result is that developers using AI-assisted development often end up with code that works but projects that are hard to reason about. Nobody is tracking what was built, why, or whether it is complete.

Romyq fills this gap. It takes a software idea, structures it into an ordered lifecycle with phases and done criteria, enforces the project constitution before execution, tracks readiness over time, and tells the operator when to stop.

• Not an AI coding assistant — Romyq does not write code itself
• Not a task manager — it manages a lifecycle, not a to-do list
• Not a SaaS platform — there are no servers, no accounts, no subscriptions
• Not cloud-dependent — everything runs locally
• Not collecting your data — zero telemetry, zero outbound calls to Romyq
• Not a wrapper around a specific AI provider — providers are configurable

Local first

All state lives in .romyq/ within your project directory. Readable, diffable, versionable. Nothing requires a network connection to Romyq.

Operator authority

Romyq advises. The operator decides. The recommendation engine tells you Continue, Pause, Review, or Stop. Execution only stops when you say so.

Transparency

Every decision Romyq makes is recorded. Every governance evaluation, task rejection, and replanning step is in the decision log. Nothing happens silently.

Honesty about limitations

The readiness score and recommendation engine are tools for estimation, not guarantees. They are as accurate as the done criteria and rules you define.

Open source

MIT licensed. The code is auditable. The project has no business model that creates incentives to collect data or lock you in.

Romyq uses AI coding assistants as execution providers. The planning provider (e.g., DeepSeek R1) decomposes the mission into phases and tasks. The execution provider (e.g., Claude Code) implements each task. Romyq coordinates between them, enforces governance, tracks readiness, and reports status to the operator.

The intended workflow: you define a mission in mission.md, run romyq init or romyq attach, thenromyq run. You steer from a second terminal withromyq steer, romyq pause, andromyq rules add. You observe readiness and recommendation in the dashboard. When the recommendation is Stop, you runromyq stop.

Romyq is in active development. The current stable release is v0.10.3. The core lifecycle engine, governance system, memory layers, readiness scoring, and recommendation engine are all implemented. Documentation and test coverage are expanding.

The project is maintained by a single developer. It has no corporate backing. Contributions are welcome. If Romyq is useful to you, consider sponsoring the project.