Backlog.md User Manual

A Markdown-native task and kanban visualization tool

Backlog.md turns any Git repository directory into a self-contained project board. Every task is a standalone Markdown file, 100% offline and private, with no internet connection required.

Core Features

  • Markdown-native: all data is stored as .md files, Git-friendly
  • AI-Ready: supports Claude Code, Gemini CLI, Codex, Kiro, Cursor, and other AI assistants
  • Dual-mode interface: terminal TUI board + modern Web browser interface
  • Rich-text paste: paste from Word or Google Docs as Markdown, with .docx file upload support
  • Dates & planning: tasks and milestones support dueDate, plannedStart, and plannedEnd; board cards show planning windows and overdue highlights in real time
  • Project health: four dimensions—due, overdue, stale, and blocked—unified in the Web statistics page and the CLI overview command
  • Wiki knowledge base: LLM-maintained incremental knowledge base, human-readable and editable; browser supports file-tree navigation, in-place editing, labels, and live sync; built-in skill installable to Claude/Codex
  • Task graph: knowledge graph of task relationships with Cypher queries (CLI backlog graph and MCP graph tools) and Web visualization; default RyuGraph WASM backend with zero native dependencies
  • Office document preview: preview .docx / .pdf / .pptx / .xlsx documents in-app; Word and Excel support editing and write-back to source files; unsupported formats show a conversion hint
  • Live SQL query blocks: embed sqlite-query / duckdb-query / mdb-query fenced blocks in documents; Web reading view renders read-only live query results (SQLite / DuckDB / Access, 500-row limit) with in-place editing
  • Chart blocks: draw flowcharts and data charts with mermaid / echarts fenced blocks; ECharts covers all built-in series and follows light/dark themes; query results can feed charts directly, or reference named query blocks via fromQuery
  • Quick notes (Memos): lightweight capture inbox under backlog/memos/; terminal / Web / MCP / search share the same storage; entity IDs and wikilink automatically connect to the knowledge web
  • Cross-platform: macOS, Linux, Windows

Two Usage Paths

  1. MCP spec-driven (recommended): AI agents manage tasks directly through the MCP protocol, no manual commands needed
  2. Manual CLI mode: users operate tasks, boards, and documents through terminal commands

Installation

npm i -g @kuwork/backlog.md

Other installation options:

  • bun add -g @kuwork/backlog.md
  • brew install backlog-md (upstream formula)
  • nix run github:MrLesk/Backlog.md (upstream flake)

Quick Start

# Initialize a project
backlog init my-project

# Create a task
backlog task create "Implement user login"

# View the TUI board
backlog board

# Start the Web interface
backlog browser

This manual covers the complete guide from basic installation to advanced AI integration.

results matching ""

    No results matching ""