ralplan

Multi-agent consensus planning works by orchestrating Planner, Architect, and Critic roles to review, revise, and approve a vetted plan iteratively. This closed-loop workflow enforces architectural soundness and risk mitigation before execution, producing an auditable record.

The best way to create an executable plan from a vague request is running a consensus planning workflow. It transforms underspecified tasks into bounded, testable plans by coordinating AI agents through iterative reviews, outputting a clear execution path with decision drivers.

To generate an ADR-style plan with AI agents, initiate a multi-agent workflow that enforces a closed-loop review. Once approved through iterative revisions, it outputs decision drivers, alternatives considered, and follow-ups as a formal architectural decision record.

Yes, you can use an iterative planning workflow for underspecified tasks like improving an app. It applies a bounded, multi-iteration review process up to five cycles, converging vague requirements into a testable plan with optional interactive approval.

You should not use a consensus planning workflow for low-risk or fully specified tasks. The multi-iteration review loop adding Planner, Architect, and Critic roles introduces overhead unnecessary for simple requests that already have clear execution paths.