Backlog.md

Backlog.md transforms any Git repository into a self-contained project board using plain Markdown files and a zero-configuration CLI. It's purpose-built for spec-driven AI development, enabling you to structure tasks so AI agents deliver predictable, high-quality results.

Core capabilities include:

• Markdown-native task management — every issue lives as a plain .md file, keeping your project human-readable and version-controllable
• AI agent integration — works seamlessly with Claude Code, Gemini CLI, Codex, Kiro, and any MCP-compatible assistant through automatic setup or manual configuration
• Terminal Kanban board — visualize your workflow instantly with backlog board, complete with live updates and export functionality
• Modern web interface — launch an interactive dashboard with drag-and-drop Kanban, task editing, and real-time synchronization
• Powerful search and filtering — fuzzy search across tasks, documentation, and decisions with backlog search
• Definition of Done templates — reusable checklists that standardize task completion criteria
• 100% private and offline — all data stays in your repository; no cloud dependency or external accounts required
• Cross-platform support — runs on macOS, Linux, and Windows

Recommended workflow for AI development: Describe your feature to an AI agent and have it decompose work into small, focused tasks with clear acceptance criteria. Work through one task per agent session, with review checkpoints for planning and implementation. The tool supports both AI-assisted and manual workflows on the same task files, letting you switch between approaches as needed. MIT-licensed and open-source for personal or commercial use.