LoopTroop: Local Multi-Model Code Agent
liviux · reddit · 2026-07-14
The author introduces LoopTroop, a local open-source GUI application they built, designed to prevent context "pollution" from logs and redundant information when running repository-level tasks.
Its workflow includes:
- LLM Council: Multiple models independently write PRDs and task breakdowns, then vote anonymously to select the best plan;
- Atomic Beads: Tasks are sliced into minimal units and executed step-by-step;
- Ralph Loops: On failure, instead of feeding long error logs back into the context, it writes a short summary, resets the worktree, and retries with a clean context;
- Human-in-the-Loop: Manual approval is retained at stages like interviews, PRDs, planning, and final diffs.
The project runs locally, connects to local git repositories, and allows customization of the council and execution models. The author also provides a 2.5-minute demo video and a GitHub repository under the MIT license.
More from coding & agent
- Codex helps build Valdiluce, an open-world game with climbing, gliding and gondolas — Dimillian · 2026-07-22
- HeyGen adds a media-sourcing skill for coding agents with 75k images and 10k tracks — HeyGen · 2026-07-22
- Agent search bottlenecks are now about variance, not raw latency — rohanpaul_ai · 2026-07-22
- LangSmith adds tracing for Pipecat, LiveKit, OpenAI Realtime, and Gemini Live — LangChain · 2026-07-22
- An MCP server signs every AI agent tool call into a verifiable Merkle chain — Funky_Chicken_22 · 2026-07-22
- Annotated transcript of a Claude Code team interview is now available — trq212 · 2026-07-22