Beads

Beads

โ˜… 4.4 ๐Ÿ”– 7000 saves ๐Ÿ’ฐ Free GitHub

Beads is a distributed, graph-based issue tracker built specifically to give AI coding agents a persistent, structured memory for their work. Its command-line tool, bd, replaces the ad-hoc markdown plans and TODO lists that agents usually keep with a real dependency-aware graph of issues that survives across sessions and machines. With about 28,000 stars on GitHub, Beads has become a popular foundation for long-running agentic coding workflows.

Overview

Traditional issue trackers are designed for humans and web UIs; agents need something they can query and update from the command line, in JSON, without merge conflicts. Beads stores issues in a Dolt database with git-native synchronization, so the same data can live in a repository and be shared between agents and teammates. Issues form a graph with explicit dependencies, use hash-based IDs so two agents working in parallel never collide, and support compaction, a form of memory decay that summarizes old closed work to keep context lean. Setup helpers wire Beads into popular agents such as Claude Code, Codex and Cursor, and the graph model supports multi-agent and multi-repository projects out of the box.

Key Features

  • Dependency-aware issue graph designed as long-term structured memory for coding agents
  • Hash-based issue IDs that eliminate merge collisions when agents work in parallel
  • Dolt-powered storage with git-native sync, so issue data travels with the repository
  • Compaction and memory decay that summarize completed work to keep agent context small
  • Multi-agent and multi-repository support, with JSON output for automation
  • One-command setup integrations for Claude Code, OpenAI Codex, Cursor and other agents

๐Ÿ’ณ Pricing

Beads is completely free and open source under the MIT license. There is no paid tier; you install the CLI and run it against your own repositories and infrastructure.

Who Is It For

Beads is for developers running AI coding agents on real projects, especially multi-session or multi-agent work where plans, blockers and progress need to persist in a structured form instead of being re-explained to the model every time.

๐Ÿš€ Visit Beads Official Site

๐Ÿšฉ Report an issue