Introducing Lutz Mode

Lutz Mode is Brokk’s new default workflow: a disciplined research → plan → build loop that keeps large language models doing the job you actually want. You describe the problem in Instructions, click the blue Lutz button, and Brokk will dig into your codebase with agentic search. Findings land in your Context as files, call sites, summaries, and notes. From this Context, Brokk proposes an explicit task list that you can split, merge, reorder, or delete. When the plan looks right, click again and Brokk builds: drop unnecessary context, research exactly what the next task needs, write code, compile, run focused tests, and feed any failures straight back into the loop for an immediate fix.

Why this is Different

Brokk makes both task list (the big picture) and context (the AI's working memory) explicit, so you always see what the model sees and you can course-correct at the first sign of confusion. Lutz Mode orchestrates Search, Code, and Architect agents so that each model has exactly what it needs to do its job.

Lutz Mode integrates Brokk’s Edit Loop, including running targeted tests, so that each task converges before moving on. (Occasionally it is necessary to write code that will not cleanly pass tests; Brokk handles this automatically, too.) And Brokk's first-class support for Git worktrees means that you can easily have Lutz Mode power through multiple concurrent tasks without stepping on its own toes.

Give Lutz a Try

Lutz Mode is context-disciplined and built to scale with real-world mess, not just vibe-coded React apps. Download Brokk and try it today!