MrLesk/Backlog.md

What it solves

Backlog.md addresses the bottleneck of human attention when working with AI agents. While agents can generate vast amounts of code quickly, reviewing thousands of lines of diffs is overwhelming. This tool provides a structured, Markdown-native way to manage tasks and acceptance criteria, ensuring humans can review the specification and implementation plan before the code is written, and review small, manageable PRs after.

How it works

It turns any folder or Git repository into a project board by storing tasks as plain .md files. It provides a zero-config CLI, a terminal-based Kanban board, and a local web UI for visual management.

For AI agents, it integrates via CLI instructions or the Model Context Protocol (MCP), allowing agents to read workflow guidelines, decompose ideas into tasks, and write implementation plans directly into the Markdown files. This creates a "spec-driven" loop: review the spec $\rightarrow$ review the plan $\rightarrow$ review the code.

Who it’s for

Developers and teams using AI coding assistants (like Claude Code, Gemini CLI, Codex, or Cursor) who want to maintain strict control over the agent's output and avoid "context window" bloat by breaking work into small, verifiable tasks.

Highlights

• Markdown-native: All tasks are stored as plain files in the repo, making them version-controlled and human-readable.
• AI-ready: Built-in support for MCP and CLI instructions for seamless agent integration.
• Local-first: No external servers or accounts; everything stays in your local filesystem.
• Visual Tools: Includes both a terminal Kanban board and a drag-and-drop web interface.
• Verifiable Scope: Supports acceptance criteria and project-wide "Definition of Done" checklists for every task.