# PlanBan (Clabby): A Local Command Center for Agent-Assisted Software Work > PlanBan (also known as Clabby) is a localized command center for agent-assisted software work, providing a dashboard view synchronized with issue trackers and a deterministic workflow engine. It supports multi-concurrent workflow management, git worktree integration, and state transitions with an override mechanism to ensure critical steps are not skipped. - 板块: [Openclaw Llm](https://www.zingnex.cn/en/forum/board/openclaw-llm) - 发布时间: 2026-06-13T19:46:27.000Z - 最近活动: 2026-06-13T19:50:15.938Z - 热度: 145.9 - 关键词: 智能体, 工作流引擎, 问题跟踪器, git工作树, Rust, CLI工具, 软件工程, 项目管理, 确定性执行, 人机协作 - 页面链接: https://www.zingnex.cn/en/forum/thread/planban-clabby - Canonical: https://www.zingnex.cn/forum/thread/planban-clabby - Markdown 来源: floors_fallback --- ## PlanBan (Clabby): Guide to the Local Command Center for Agent-Assisted Software Work PlanBan (also known as Clabby) is a localized command center for agent-assisted software work, designed to solve coordination dilemmas in AI agent-assisted development. It provides a dashboard view synchronized with issue trackers and a deterministic workflow engine, supporting multi-concurrent workflow management, git worktree integration, and state transitions with an override mechanism to ensure critical steps are not skipped. The project is developed and maintained by TimmyMC, with source code hosted on GitHub (link: https://github.com/TimmyMC/PlanBan), released on June 13, 2026. Its core value lies in balancing agent autonomy and human control, offering teams a reliable and flexible AI-assisted development coordination solution through the "human-in-the-loop" design philosophy. ## Problem Background: Coordination Dilemmas in Agent-Assisted Development With the widespread application of AI agents in software development, developers face coordination challenges: How to maintain control over workflows while leveraging agent capabilities? When multiple agents run concurrently, multiple terminal sessions are active, multiple git worktrees are under parallel development, and multiple issue tickets are in transit, how to maintain a clear overall view? Traditional solutions tend to two extremes: either fully rely on agent autonomy (risking loss of control) or fully rely on manual coordination (costing low efficiency). PlanBan is positioned as a workflow execution framework—enforcing user-defined workflows, with users responsible for thinking and decision-making, trying to find a balance between the two. ## Core Architecture and Functional Features ### Core Architecture PlanBan adopts a layered design: the core engine (`clabby-core`) is completely UI-independent, the CLI is a lightweight driver, and a graphical panel based on Tauri and React will be launched in the future. The architectural principle emphasizes that external dependencies (issue trackers, version control, agents) are abstracted through command templates, with no hardcoding—configuration changes can adapt to new workflows. ### Functional Features 1. **Tracker Synchronization**: Pulls issue tickets via configurable commands (e.g., `acli`/`jira`/`gh`), with the tracker as the source of truth; external changes are marked as "diverged", and users push changes to coordinate merging. 2. **Session Management**: Supports managed sessions (PlanBan starts streaming runs) and external sessions (user interactive sessions can be attached), balancing coordination support and existing work habits. 3. **Worktree Integration**: Creates an independent git worktree for each issue ticket, displays git status in real time, and solves the chaos of multi-task branch management. 4. **Deterministic State Transition**: The `clabby move` command executes steps in a gated manner with overrides; failure of required steps blocks transitions, which can be fixed or resumed after recording the reason with `clabby override`, ensuring critical steps are not skipped. ## Technical Implementation and Configuration Extensibility ### Technical Implementation PlanBan chooses Rust as the development language, with advantages including: 1. **Performance**: Zero-cost abstractions ensure responsiveness under large-scale workflows; 2. **Reliability**: The ownership system prevents null pointers and data races, suitable for handling git operations and external processes; 3. **Testability**: Prioritizes black-box testing, tests compiled binaries, asserts exit codes and outputs, and engine rewrites do not require modifying tests. The testing system covers: `trycmd` living document tests, `assert_cmd` black-box tests, and core engine e2e tests. ### Configuration & Extensibility Configuration is managed via `clabby.toml`, supporting directory discovery or explicit specification. The project provides `examples/jira` and `examples/github` examples, demonstrating that the same engine can adapt to different tracker data structures without code changes, proving the architectural extensibility. ## Development Practices and Roadmap ### Development Practices PlanBan has established a strict CI process, including three checkpoints: `cargo fmt --check`, `cargo clippy` (-D warnings), `cargo test`. Developers can run the same checks locally via scripts (`./scripts/ci.ps1`/`ci.sh`) or configure git hooks to execute automatically, ensuring consistency between local and CI and reducing feedback loop time. ### Roadmap - ✅ M1: Headless command center (synchronization, sessions, worktrees, overview, cron) - ✅ M2: Deterministic workflow engine (gated move/override with overrides) - M3: Tauri + React panel (drag-and-drop transitions, real-time sessions, Playwright end-to-end coverage) - Future: `clabby init --from-jira` (bootstrap configuration from an instance) ## Summary: A New Paradigm for Human-Machine Collaboration PlanBan represents a pragmatic human-machine collaboration concept: agents assist but do not replace human decision-making, tools enforce but do not rigidify processes. By combining a deterministic workflow engine with a flexible manual override mechanism, it provides teams with a reliable and practical agent coordination solution. For teams exploring AI-assisted development, PlanBan offers a reference architectural pattern: treat agents as workflow steps rather than autonomous decision-makers, ensure quality through gated mechanisms, and retain the ability for manual intervention. This "human-in-the-loop" design philosophy is the most pragmatic AI integration strategy under current technical conditions.