
Beads
Git-backed issue tracker for AI agents — persistent memory across sessions.
About
Beads by Steve Yegge is an issue tracker built for AI-supervised coding workflows. Issues live as JSONL in .beads/ inside your repo and sync through git, giving agents persistent, mergeable memory across sessions and parallel workers — fixing the '50 First Dates' problem where every session starts blank. Dependencies are first-class (blocks, depends-on, parent-child), and bd update --claim gives multi-agent claim semantics so concurrent agents don't grab the same task. The bd CLI is tool-agnostic and complements Claude Code's in-session tasks with durable cross-session state.
The honest picture: it's genuinely useful for long-running agent work, but it adds process overhead — Yegge's own guidance says keep the DB small (under ~200–500 issues) and run bd doctor regularly. A Rust rewrite (br) and a Tauri desktop app exist for humans who want visibility into what agents are doing. MIT licensed.
What it can do
- Git-backed issue DB (JSONL in .beads/)
- Dependency graphs: blocks, parent-child, discovered-from
- Multi-agent claim semantics (bd update --claim)
- Formulas and molecules for workflow templates
- bd ready queue surfacing unblocked work
- Works offline; syncs with git
Setup
brew install beads (or build from source), then bd init in your repo. Agents use bd create/list/show/update/close with --json; run bd sync at session end.
Prerequisites
A git repository; the bd CLI in PATH.
Use cases
Works with
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 46; Trending score 80/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.