Do I need to initialize Beads manually?

Yes, a human must run 'bd init' once in the repository to set up the .beads/ directory before the AI agent can begin tracking tasks.

When should I use Beads instead of TodoWrite?

Use Beads for work that spans multiple sessions, has complex blockers, or requires long-term context; use TodoWrite for simple, linear tasks that will be finished in the current session.

What happens to my tasks when the AI conversation is compacted?

Tasks remain safely stored in the local .beads/ directory, allowing the AI to re-read the full state and history in the next session without losing progress.

Can Beads handle multi-developer collaboration?

Yes, because it is git-backed, you can use 'bd sync' to push and pull task states, allowing multiple people and AI agents to work from the same issue graph.

How does Beads differ from standard markdown task lists?

Unlike markdown lists, Beads tracks complex dependencies and stores data in a structured git-backed database that survives when the AI's conversation memory is cleared or compacted.

Beads - Persistent Task Memory

Name: Beads - Persistent Task Memory
Author: edmundmiller

byedmundmiller

•

生产力与工作流

Manages complex, multi-session tasks and dependencies with a git-backed graph-based issue tracker that survives conversation compaction.

Beads (bd) provides a robust, persistent memory system for Claude, allowing it to track intricate work across multiple sessions and compaction cycles. Unlike transient task lists, Beads uses a dependency-aware graph stored directly in your git repository, ensuring that context, blockers, and progress notes remain accessible even after long breaks or history resets. It is the ideal tool for exploratory research, multi-stage feature development, and any project where maintaining a clear audit trail of decisions and dependencies is critical for AI-human continuity.

主要功能

01Smart task discovery prioritizing ready work and identifying project bottlenecks

02Automatic synchronization with remote repositories for collaborative AI-human workflows

03Git-integrated persistence that survives AI conversation history compaction

04Graph-based dependency tracking to manage blockers and parent-child task relationships

05Detailed audit trails including status changes, notes, and decision history

0642 GitHub stars

使用场景

01Maintaining a shared task state between human developers and AI agents within a single repository

02Tracking exploratory research where new sub-tasks are discovered incrementally

03Managing large-scale software refactors that span several days and multiple chat sessions

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add edmundmiller/dotfiles beads

For use in Claude.ai and ChatGPT