beads

Provides persistent multi-session memory for LLM agents using a Dolt-backed store.

Updated Aug 23, 2026

Search Tags:#dependencies #beads #multi-session #compaction #persistent-memory #agent-work

Ready to run

npx skills add https://github.com/HuynhSang2005/delivery-app --skill beads-huynhsang2005

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: beads
Source: https://github.com/HuynhSang2005/delivery-app/tree/main/.agents/skills/beads
Command: npx skills add https://github.com/HuynhSang2005/delivery-app --skill beads-huynhsang2005

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Beads provides persistent task memory for AI agents, allowing long-running work to survive conversation compaction by recording decisions, context, and dependencies in a Dolt-backed store.

Core Features & Use Cases

Persistent memory across sessions for multi-step work and complex collaborations
Dependency tracking with blocks, related, and discovered-from relationships to coordinate effort
Session handoff and resume workflows to recover progress after gaps
Progressive disclosure of concept content via resources and patterns for advanced usage

Quick Start

Run bd ready to see available work, then claim an issue to begin.

Frequently Asked Questions about beads

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I maintain persistent task memory for AI agents across multiple sessions?▼

Persistent task memory for AI agents is maintained by recording context, decisions, and dependencies in a Dolt-backed store, allowing long-running work to survive conversation compaction and resume smoothly across sessions.

What is the best way to track task dependencies and relationships during complex AI workflows?▼

Tracking task dependencies during complex AI workflows is done by defining blocks, related, and discovered-from relationships in a persistent memory store to coordinate effort and ensure structured handoffs.

Why does my AI agent lose progress and context after a conversation compaction?▼

AI agents lose progress after conversation compaction because context is not persisted externally; using a Dolt-backed memory store solves this by explicitly recording decisions and dependencies for cross-session continuity.

Can I resume a multi-step AI agent task after closing the session?▼

Yes, you can resume a multi-step AI agent task after closing the session by using session handoff and resume workflows that recover progress from the persistent memory store. Run bd ready to see available work.

Do I need a Dolt database to enable long-horizon planning for AI agents?▼

Yes, a Dolt-backed store is required to enable long-horizon planning, providing the persistent storage needed to retain knowledge and dependencies for collaborative workflows across multiple sessions.