# Beads — Git-Backed Dependency Graph Issue Tracker Built for AI Coding Agents

> A dependency-aware issue tracker built for AI coding agents — versioned graph storage with hash IDs that survive multi-agent merges.

- Org: Gastown
- Date: 2026-04-24
- Category: repo, tool
- Source: https://github.com/gastownhall/beads
- Canonical page: https://ai-tldr.dev/releases/beads-agent-issue-tracker/

## In short

Beads replaces markdown planning files with a Dolt-backed dependency graph for coding agents. Hash IDs prevent multi-agent merge conflicts; semantic memory decay keeps context windows lean; `bd prime` auto-injects ready tasks at session start. 21.6k GitHub stars.

## What it is

Beads (bd) is an open-source distributed graph issue tracker designed specifically for AI coding agent workflows. It replaces unstructured markdown planning files with a Dolt-powered (version-controlled SQL with cell-level merging) dependency graph. Issues have typed relationships (relates_to, duplicates, supersedes, replies_to) and dependency links that auto-surface ready tasks. Agents call `bd prime` at session start to get a JSON list of next actionable items injected into their context window. v1.0.3 released April 24, 2026. 21.6k GitHub stars.

## How it works

Issues get short hash IDs (bd-a1b2) that prevent merge conflicts in multi-agent scenarios where two agents work in parallel. Closed issues are compressed with semantic memory decay — a summarization step that preserves key decisions while reducing context pressure. Hierarchical tasks support epics with subtasks (bd-a3f8.1.1). A persistent memory layer (`bd remember`, `bd recall`) survives session and account rotations. Available via Homebrew, npm, or direct scripts on macOS, Linux, Windows, and FreeBSD.

## Why it matters

Long-horizon coding tasks break down because agents lose context across sessions, and multi-agent workflows create conflicting state. Beads addresses both: the Dolt backend handles concurrent writes without conflicts, and semantic decay keeps relevant context available without bloating the context window. The 21.6k stars in a relatively new project signal strong practitioner adoption for agentic development infrastructure.

## Who it's for

Developers running long-horizon coding agents or multi-agent workflows with Claude Code, Codex, or similar tools

## Try it

brew install beads && bd init

## Links

- [Repo](https://github.com/gastownhall/beads)
- [Releases](https://github.com/gastownhall/beads/releases)