AI Decision System
Replayable, offline-verifiable records for AI-assisted go/no-go decisions: experiment promotions, typed model answers and human-approved actions.
翻譯準備中
尚未偵測到更新。GitHub 發布版本後會顯示在這裡。
個人 GitHub 帳號下公開儲存庫的擁有者可以登入並驗證所有權,自行管理介紹,無需管理員核准。
登入並產生驗證檔案,在 30 分鐘內將其提交至此儲存庫的預設分支。其他分支或分叉儲存庫中的檔案無效。然後在下方檢查檔案,即可無需管理員核准直接管理專案。請保留檔案以維持管理權限。
如需登入,將前往登入頁面。登入後請回到此處,再次點擊所選的驗證按鈕。
在專案資料夾中執行 /nomorevibe verify,即可驗證所有權並自行管理介紹。
一分鐘完成驗證Local-first control plane for multi-agent AI development — Claude + Gemini with a live Kanban dashboard
翻譯準備中
取自儲存庫 README
正在載入完整 README。 以下為已儲存的摘錄。
The local-first control plane for multi-agent AI development.
See it in action: Brainstorming → Planning → Automated Implementation
LaneConductor is a sovereign developer environment that orchestrates AI agents (Claude, Gemini, and more) across multiple repositories, giving you real-time visibility through a unified Kanban dashboard — with zero cloud dependency and zero cost.檢視完整 README ↗The local-first control plane for multi-agent AI development.
LaneConductor is a sovereign developer environment that orchestrates AI agents (Claude, Gemini, and more) across multiple repositories, giving you real-time visibility through a unified Kanban dashboard — with zero cloud dependency and zero cost.
conductor/tracks/ as the source of truth, enabling agents and humans to coordinate seamlessly.While standalone terminal tools (such as Claude Code, Gemini CLI, Aider, or Cursor) provide interactive single-session coding, LaneConductor acts as an autonomous multi-agent control plane that orchestrates complex development lifecycles across your entire repository.
| Dimension | Standalone Coding Agents | LaneConductor Control Plane |
|---|---|---|
| Architecture | Ephemeral, single-terminal REPL | Sovereign multi-agent control plane with background worker fleet |
| Visibility | Terminal scrollback buffer | Real-time Kanban dashboard backed by local Postgres & Markdown bus |
| Workspace Isolation | Runs directly in your active working tree; prone to dirty checkouts | Git Worktree Isolation: automated worktrees per track keep main pristine |
| Lifecycle & Safety | Open-ended chat; manual testing & manual git committing | Structured Conductor Pipeline: Plan → Implement → Review → Automated Quality Gates → PR/Merge |
| Target Switchability | Single monolithic conversation session | Switch between The Manager (instance health & gap triage) and Track Workers |
| Steerable Alignment | Ad-hoc chat disconnected from project roadmaps | Conversational steering updates live design docs (spec.md, plan.md) and re-queues tracks |
| Parallel Execution | One task at a time; blocks your terminal | Multi-track concurrent queues running headlessly in parallel worktrees |
conductor/tracks/, serving as a durable, human-and-agent readable single source of truth.Every Track folder under conductor/tracks/ is controlled by a stateful index.md file. You can configure execution and workflow behavior per-track using simple markers:
**Workspace**: main|branch):
branch (default): Isolation in a dedicated git worktree on a track branch.main: Execution directly in the primary checkout (perfect for local developer-agent pairing or self-hosted worker infrastructure updates).**Merge Mode**: pr|direct):
pr (default): Automatically pushes the branch and opens a GitHub Pull Request upon passing the Quality Gate. You can review, preview, and merge it straight from the dashboard.direct: The worker merges the track branch directly to main without opening a PR.**Auto Run**: yes|no):
yes: Allows the worker to claim the track from the queue and execute it automatically through planning, implementation, and review.no (or omitted): Restricts execution to manual trigger actions from the UI or CLI.# 1. Install the lc CLI globally
git clone https://github.com/meller/laneconductor.git && cd laneconductor && make install
# 2. Initialize in your project
cd ~/your-project && lc setup
# 3. Start the worker and dashboard
lc start && lc ui
The Kanban dashboard will be available at localhost:8090.
No CLI, no local DB, no UI required. Works on Windows, macOS, and Linux.
laneconductor folder from .claude/skills/ into your Claude Desktop skill directory./laneconductor setup
/laneconductor newTrack "Feature Name" "Short Description"/laneconductor plan NNN/laneconductor implement NNN
Check conductor/tracks.md at any time for a project-wide summary.Use the Claude Code skill to automatically scan your codebase and generate project documentation:
# Within Claude Code
/laneconductor setup scaffold
For a deep dive into operating modes, CLI reference, and workflow configuration, visit the LaneConductor Knowledge Base.
bin/: The universal lc command-line tool.conductor/: Core orchestrator logic and heartbeat worker.ui/: Vite + React Kanban dashboard and Express API.cloud/: Firebase functions and Postgres connectors for remote/team mode.migrations/: Schema versioning using Atlas..claude/skills/: The specialized AI skill definition for Claude.This project is licensed under the Elastic License 2.0 - free to use, modify, and self-host; the one restriction is offering it to third parties as a hosted or managed service.