Local AI orchestrator for careful, repository-scale code delivery
Experience a blend of multi-model planning, iterative automated repair, and isolated shipping: LoopTroop breaks tickets into planning, execution, and worktree-based shipping. Developed by LoopTroop AI, this Windows-oriented utilities tool targets professional engineers and researchers who need high-precision, repository-scale AI assistance. Key capabilities include an LLM Council for consensus planning, a Ralph-loop for automated test-and-fix cycles, and OpenCode worktree isolation to keep the main repository untouched until human approval.
What the tool actually does for complex engineering tasks
LoopTroop orchestrates multi-file development by splitting a ticket into staged beads that move from a council-written plan to isolated implementation. The app uses an LLM Council to draft and vote on technical plans, then runs each bead through the Ralph-loop recovery system for iterative testing and fixes before creating an OpenCode worktree for the change. This design scopes work at repository scale and enforces review gates before merging.
How it interacts with a developer's system during operation
The tool runs a local-first GUI in the browser while a local backend manages state in SQLite and YAML artifact files, so persistent state stays on the machine. It requires Node.js 24+ and Git, which the app invokes for worktree creation and commits. The app isolates edits in temporary worktrees, preventing direct writes to the main workspace until the developer approves a pull request.
Safety, verification, and human-in-the-loop controls
Safety is implemented through mandatory human-gated approval workflows and OpenCode worktree isolation: AI-generated edits occur in temporary worktrees and only merge after manual review. The LLM Council phase provides cross-model consensus to reduce single-model bias. Durable local state and artifacts make it possible to audit decisions and reproduce the planning sequence for compliance or postmortem analysis.
Who can operate the tool and what technical knowledge it expects
The app targets professional engineers and AI researchers rather than casual users; prerequisites list Node.js, Git, and a configured AI model provider, so initial setup requires developer skills. In addition, the human-gated workflow and multi-model planning assume reviewers can read design documents and diffs. The interface is local-first and browser-based, but repository-scale tasks and council tuning require technical involvement.
Strong option for teams needing auditability, slower correctness, and controlled AI changes
LoopTroop suits engineers who accept a deliberate, verifiable AI workflow: it provides multi-model planning, iterative automated repair, and enforced worktree isolation to protect the main repository. Expect a steeper setup and an operational workflow that assumes reviewer discipline; teams wanting automatic, conversational speed will find its "slow and perfect" posture unsuited to quick fixes.
